Space Weather Effects on the Middle and Upper Atmosphere by Joshua M

Space Weather Effects on the Middle and Upper Atmosphere by Joshua M. Pettit B.S., Cornell University, 2012 M.S., University of Colorado Boulder, 2014

A thesis submitted to the Faculty of the Graduate School of the University of Colorado in partial fulfillment of the requirements for the degree of Doctor of Philosophy Department of Atmospheric and Oceanic Sciences

2019

This thesis entitled: Space Weather Effects on the Middle and Upper Atmosphere written by Joshua M. Pettit has been approved for the Department of Atmospheric and Oceanic Sciences

______

Cora E. Randall

______

V. Lynn Harvey

Date ______

The final copy of this thesis has been examined by the signatories, and we Find that both the content and the form meet acceptable presentation standards Of scholarly work in the above mentioned discipline.

Pettit, Joshua M. (Ph.D., Atmospheric and Oceanic Sciences) Space Weather Effects on the Middle and Upper Atmosphere Thesis directed by Prof. Cora E. Randall

Abstract:

Impulsive solar events (ISEs) have a pronounced impact on the middle and upper atmosphere. Solar flares emit copious amounts of x-ray and EUV radiation that can ionize the D, E, and F regions of the ionosphere. Coronal mass ejections and geomagnetic storms can cause energetic particle precipitation. This creates nitrogen oxides and hydrogen oxides, which catalytically destroy ozone. Most global climate models exclude both radiation from solar flares and particles from geomagnetic storms and as a result, does not capture the atmospheric effects from these events.

The work presented here improves modeling of ISEs by including high-energy photons from solar flares and energetic electrons from geomagnetic storms in the Whole Atmosphere Community Climate Model (WACCM). This is necessary in order to simulate accurately chemistry changes during and after ISEs.

We found significant impacts on the D and lower E-regions of the ionosphere from WACCM simulations of the September 2005 solar flares. Effects from the flares were confined to the dayside of Earth but persisted a few days below the thermosphere. Large increases in electron densities were found from the lower D-region through the top of the model (~140 km). Moderate production of odd nitrogen in the mesosphere and lower thermosphere occurred during the larger flare events of the time period. We showed that differences in flare spectrum had a distinct impact on the intensity, length and location of the effects.

This investigation also targeted the influence of electron precipitation during the 2003 SH polar winter. Two different energetic electron data sets were included in simulations using the WACCM model with D-region chemistry. Both electron data sets contained electrons >30 keV using the Medium Energy Proton and Electron Detector (MEPED), however only one uses both telescopes onboard MEPED. Both simulations that included electrons >30 keV compared better with observations than the baseline simulation. The simulation that included both telescopes

iii agreed better with observations than the other simulations. The results confirm that including both telescopes during geomagnetically active time periods leads to better agreement with observations in the precipitation effects, particularly at sub-polar latitudes.

Acknowledgements

Thank you to Cora E. Randall for the immense amount of time and effort she spent attempting to turn me into a productive scientist. Thank you to V. Lynn Harvey for her help in improving my scientific programming. Thank you to Jeff France, Ethan Peck, Laura Holt, Susanne Benze, Matthias Brakebusch, Adrianna Hackett, Colby Brabec, and Ryan Ligon for sitting through endless group meetings listening to me ramble. I would also like to thank my close friends and all my old roommates for trying to keep me grounded over the past few years.

Thank you to Karlsruhe Institute for Technology for their contribution of MIPAS data. Thank you to Phil Chamberlin for access and support of the FISM data. This research has been funded by the NASA Living With a Star program, grant NNX14AH54G, the NSF Frontiers of Earth System Dynamics program, grant AGS 1135432, and the NSF Coupling, Energetics and Dynamics of Atmospheric Regions, grant AGS 1651428. The CESM project is supported by the National Science Foundation and the Office of Science of the U.S. Department of Energy. NCAR is sponsored by the National Science Foundation. The NASA High-End Computing (HEC) Program through the NASA Advanced Supercomputing (NAS) Division at Ames Research Center provided resources supporting this work.

Table of Contents Chapter 1. Introduction ...... 1 1.1 Goals and Motivation ...... 1 1.2 Energetic Particle Sources ...... 2 1.3 Particle Precipitation ...... 12 1.4 Middle Atmosphere Chemistry ...... 22 1.5 Middle Atmosphere Meteorology ...... 30 1.6 Whole Atmosphere Community Climate Model ...... 34 1.6.1 Model Description ...... 34 1.6.2 Ionization from protons ...... 36 1.6.3 Ionization from electrons ...... 37 1.6.4 Ionization from photons ...... 42 2. Effects of the September 2005 Solar Flares and Solar Proton Events on the Middle Atmosphere in WACCM ...... 46 2.1 Introduction ...... 46 2.2 Model Description...... 52 2.3 Numerical Experiments ...... 54 2.4 Results and Discussion ...... 57 2.5 Conclusions ...... 57 3. Atmospheric Effects of >30 keV Energetic Electron Precipitation in the Southern Hemisphere Winter of 2003 ...... 76 3.1 Introduction ...... 76 3.2 Model and Datasets ...... 80 3.3 Results and Discussion ...... 87 3.4 Conclusions ...... 102 4. Overall summary and conclusions and future work ...... 106 Bibliography ...... 111 Appendix ...... 128 Appendix A List of Acronyms ...... 128

Tables Table

1.1 HOx Production from Ionization in WACCM ...... 37 2.1 WACCM Model Simulations ...... 54

vii

Figures Figure 1.7 Schematic of Sun’s Interior ...... 4 1.8 Sunspot butterfly diagram and sunspot area ...... 5 1.9 Schematic of charged particle motion ...... 11 1.10 Schematic of ionization profiles of particles ...... 14 1.11 Disposition altitudes of XUV and EUV energy ...... 23 1.12 Lifetime of NO ...... 30 1.13 Middle atmosphere zonal wind circulation ...... 32 1.14 Middle atmosphere meridional wind circulation ...... 33 2.1 SPE ionization production rate ...... 55 2.2 FISM 1.3 nm irradiance ...... 56 2.3 Flare irradiance from FISM at several wavelengths ...... 57 2.4 Simulated electron density background and differences ...... 59

2.5 Simulated temperature, electron, NOx changes for 9-7 flare ...... 61

2.6 Simulated temperature, electron, NOx changes for 9-9 flare ...... 63

2.7 Simulated HOx, NOx differences in NH polar region ...... 64

2.8 Simulated HOx, NOx differences in SH polar region ...... 65

2.9 Simulated HOx, NOx differences in equatorial region ...... 66

2.10 Simulated electron, NOx differences from boxed region ...... 68

2.11 Simulated HOx, NOx vertical profile differences ...... 69

2.12 Simulated O3 differences between forced and baseline runs ...... 70 2.13 Simulated NO with comparisons with ACE ...... 71 3.1 Ionization Rates at 90km and 75km for CMIP6 and MP15 ...... 88 3.2 Ionization Rates in 2003 SH winter ...... 90 3.3 Simulated OH changes from WACCM in 2003 SH winter...... 91

3.4 Simulated NOx compared with MIPAS NOx ...... 93

3.5 NOx comparisons between WACCM and MIPAS ...... 94

3.6 Simulated NOx compared with MIPAS NOx at mid-latitudes ...... 96 3.7 Solar occultation observations compared with WACCM simulations ...... 98 3.8 Simulated CO from WACCM compared with MIPAS CO ...... 102

viii

Chapter 1

1: Introduction

1.1. Goals and Motivation

The overarching goal of this work is to improve our understanding of the atmospheric effects of space weather events, including solar flares, solar proton events (SPEs), and energetic electron precipitation (EEP). The work is motivated by evidence that these phenomena have pronounced effects not only on the ionosphere and near earth region where satellites exist and astronauts work, but also on the chemistry and dynamics of the atmosphere below these regions.

Yet recent investigations suggest that global climate models are inadequate to explain fully the effects of space weather on the middle and lower atmosphere, preventing us from comprehensively simulating previous events and predicting the effects of future events. The work in this dissertation addresses this shortcoming.

Including in global climate models the layers of atmosphere above the troposphere enables a comprehensive earth system approach to climate modeling. This requires treatment of space weather effects on the middle and upper atmosphere, since such effects are thought to instigate nonlinear coupling with the atmosphere below (e.g., Callis et al., 1998a; 1998b; 2001;

Rozanov et al., 2005; Seppälä et al., 2009). Current theories show the middle atmosphere

(stratosphere and mesosphere) is expected to cool while the troposphere warms (Solomon et al.,

2018). This is due to carbon dioxide radiatively cooling the atmosphere above the tropopause.

However, space weather events such as particle precipitation and solar flares can, in principle, also affect middle atmosphere temperatures because of effects on ozone, a radiatively active gas

(e.g., Rozanov et al., 2012). Thus, precisely quantifying anthropogenic effects on the middle atmosphere requires distinguishing between any temperature changes caused by human-induced increases in CO2 and those caused by naturally varying space weather events. In addition, it is necessary to simulate any feedback processes, whereby anthropogenic changes to the middle atmosphere might alter the response of the atmosphere to space weather events.

This dissertation describes the first known inclusion of solar flares into a fully coupled global climate model, and the resulting effects on the middle and upper atmosphere. It also describes the inclusion of a new EEP data set, with comparisons between the model simulations and observations of the mesosphere and stratosphere during and after geomagnetic storms. Due to limitations in energetic electron observations, global climate models usually ignore their effects and as a result, underestimate chemistry changes and ozone destruction in the middle atmosphere. We show here the importance of including both the solar flare and EEP effects in model simulations. As described more fully in subsequent sections, the work here uses the

National Center for Atmospheric Research (NCAR) Whole Atmosphere Community Climate

Model (WACCM), version 4 (Gent et al., 2011; Marsh et al., 2013). WACCM is used as the atmospheric component of the Community Earth System Model (CESM; Hurrell et al., 2013).

1.2. Solar Flares and Energetic Particle Sources

Six unique layers separate the sun’s interior and atmosphere. The innermost region is the core, which is the energy source of the sun through nuclear fusion of hydrogen into helium. The

2 second layer is the radiative zone, which is surprisingly opaque, trapping photons generated from the core for thousands of years before escaping. The third region is the convective layer, which is not unlike convective regions here on Earth. The hotter temperatures at the base of this layer cause instability and buoyancy, resulting in plasma rising into the photosphere. The convection in this region causes the granular structure that we see on the photosphere of the sun (Priest,

1995). The photosphere is the fourth layer of the sun and can crudely be called the sun’s surface.

When looking at the sun in visible wavelengths, this is the region where we can see sunspots and other surface features. This region is also home to the beginning of the magnetic instabilities that result in space weather events that can affect the Earth. Magnetic flux tubes rise through magnetic buoyancy beneath the photosphere’s surface, and when they breach the photosphere, they give rise to sunspots that we know are the active regions of magnetic activity. Next, we reach the solar atmosphere, which is home to the final two layers. The first atmospheric layer is the thin chromosphere. The final layer is the solar corona. The corona is quite interesting in that it is much hotter than the sun’s surface (photosphere). Alfvѐn waves are believed to contribute to the large temperature increase seen here (Tomczyk et al., 2007). This is the region where solar wind originates and where coronal mass ejections (CMEs) launch plasma into space. Figure 1 shows an overview of the various regions of the sun just described, as well as locations of some of the basic events that we see.

Sunspots are probably the most familiar features on the sun. Early astronomers, particularly in China and Korea, took detailed notes of sunspot numbers and approximate sizes

(Clark and Stephenson, 1978; Wittman and Xu, 1987; Lean et al., 1995). However, it was not until 1843 that the sunspot cycle was discovered and the early 20th century before sunspots were linked to magnetic activity (Charbonneau, 2010). As mentioned, sunspots are one of the homes of explosive magnetic activity on the sun. The sunspot or solar cycle is often considered an 11- year cycle, but is in fact a 22-year cycle. The Hale polarity law tells us that the polarity of active regions on the sun are the same on the same hemisphere and opposite of those across the equator.

This polarity reverses during the next 11-year period, leading to the opposite polarity of active regions on the same hemisphere as the previous 11 years. Thus, it takes 22 years to return to the same polarity and position in the same hemisphere (Hale et al. 1919). The magnetic poles on the sun are also found to reverse polarity in the middle of the 22-year solar cycle (Fan, 2009). This occurs during the period referred to solar maximum, which is the period where the most sunspots

4 are visible. During the early years of solar maximum conditions, observations tend to show formation of sunspots at higher latitudes, which slowly descend toward the equator throughout the solar cycle. This is referred to as the butterfly effect or Sporer’s law (Babcock, 1961) and is shown in Figure 2.

Sunspots are cooler than the surrounding area because the magnetic field inhibits convection from below. Nonetheless, they are harbors of magnetic energy, and an active region can trap significant amounts of energy. Once the magnetic field can no longer confine the magnetic energy within the region, the active region prominence and its overlying arcade begins to rise from instability. This marks the ‘prephase’ of a solar flare event. Magnetic flux tubes rising through the photosphere allow for the buildup of magnetic energy via the shearing and twisting of plasma. Once the field strength can no longer support the energy, the stretched field lines break and then reconnect, causing an explosive release of energy (Priest, 1995). The

5 reconnection signals the ‘onset’ phase, which is sometimes referred to as the ‘impulsive phase’.

During the ‘main phase’ of a solar flare, reconnection continues, creating x-ray radiation as the field closes down (Priest, 1995). Solar flares are one of the primary causes of magnetic energy release from the sun, and can be defined as a sudden brightening of the soft x-ray spectrum as seen from Earth. The flux in the soft x-ray determines the intensity of the flare. The

Geostationary Operational Environmental Satellite (GOES) X-ray Sensor (XRS) measures the

0.1 – 0.8 nm wavelength bin using 5-minute averages. Using a logarithmic scale, classification of solar flares is determined by peaks in this wavelength bin using the letters A, B, C, M, and X, going from smallest to largest. Spikes in the flux of radiation in the hard x-ray, radio, EUV, and

UV spectral regions are also observed from solar flares (Forbes, 2010).

CMEs and solar flares are closely correlated (Zhang et al., 2001). Not surprisingly then,

CMEs are also associated with solar magnetic field activity. However, it is difficult to connect them mechanistically due to limited observations of the lower solar corona. CMEs can be described briefly as spontaneous ejections of plasma in the solar corona (Wagner et al., 1984).

The ejected plasma is primarily made up of protons, electrons and some alpha particles that are intensely energized by the release of magnetic energy. CMEs tend to move faster than the solar wind, which creates collisionless shock waves along the front of a CME. Collisionless shock waves differ from conventional gas shock waves in that collisions are not the dissipation mechanism. Instead, plasma waves create turbulence within the shock to cause dissipation

(Burgess, 1995). Shock waves play the important role of energizing the particles within the plasma that moves within the magnetic cloud of a CME. The wave interference from these shocks can energize solar particles several times before the particles leave the shock region

(Krauss-Varban, 2010). If the CME is directed toward Earth, the magnetic cloud eventually reaches the Earth’s magnetosphere. CMEs are effective at changing the Bz component of Earth’s magnetic field southward. This allows for dayside magnetic reconnection, which permits energetic particles access to the atmosphere more easily (Eastman et al., 1976; Burch 1972).

Solar particles from the CME are bound by the magnetic field and move toward the magnetic poles, descending in altitude until reaching Earth’s neutral atmosphere. When the proton monitors around earth spike to a certain level, this suggests that a solar proton event (SPE) has been triggered (Jackman et al., 2008). The energy added to Earth’s magnetosphere from a CME can also trigger geomagnetic storms, altering electron populations in the radiation belts

(Goplaswamy et al., 2007, Zhang et al., 2007).

One of the most consistent sources of energetic particles at Earth is the solar wind. Fast solar wind comes from coronal holes that typically form in the polar region of the sun. Coronal holes are common in the declining phase of solar max and are generally the cause of geomagnetic storms during low solar activity. They often align with the sun’s 27-day rotational cycle when magnetic field lines in the sun’s corona open up to space (Tsurutani et al., 1995).

This allows a direct route for particles to travel into the interplanetary medium. Coronal holes can last far longer than solar flares and CMEs, so their influence on the chemistry of the middle atmosphere could exceed that of more impulsive events (Rodger et al., 2010a). Co-rotating interactive regions (CIRs) can form as a consequence of high speed streams (often from coronal holes) that ‘collide’ with lower speed streams in the interplanetary medium (Richardson, 2003).

They often remain stable through solar rotation and can repeatedly cause geomagnetic storms here at Earth (Tsurutaniet al., 1995). Originally it was thought that CIRs rotated on the 27-day

7 solar cycle, however evidence has shown shorter modes of 9 days and 13.5 days as well (Lei et al., 2008; Lei et al., 2010). High-speed streams from CIRs last longer than CMEs because the holes remain open while the sun rotates, allowing for multi-day duration of activity (Vršnak et al., 2007). While CIRs typically do not create solar proton events (SPEs), they are an important source of geomagnetic activity (Miyoshi and Kataoka, 2005; Richardson et al., 2006; Alves et al,

2006). Fast streams can create shock waves similar to those of CMEs, accelerating particles within the solar wind and energizing the magnetosphere (Fisk and Lee, 1980). Coronal holes and

CIRs are also capable of triggering geomagnetic storms. As we will find, precipitation of electrons from the radiation belts often occur during geomagnetic storms. Therefore, CIRs can be important contributors to atmospheric chemistry changes in the middle and upper atmosphere.

Slow wind streams are often associated with streamers from equatorial latitudes of the sun.

Slower solar wind tends to have less of an impact on impulsive solar events like geomagnetic storms and SPEs (Richardson et al., 2003).

Geomagnetic storms are another space weather disturbance that can cause changes to the middle and upper atmosphere and are primarily caused by CMEs, CIRs, and coronal holes. They are created by large disturbances in Earth’s magnetosphere. These disturbances create large changes in electron populations within the outer radiation belts (Green et al., 2010). The outer radiation belts consist primarily of energetic electrons, whereas the inner belt consists mostly of protons. Variations to the inner radiation belt occur on time scales of approximately 100 days, while variations in the outer radiation belts can occur in about 1 day or shorter (Millan and

Thorne, 2007). The difference in time scale is due to the loss processes between the two regions.

Acceleration processes in the outer radiation belts occur on shorter time periods than the inner

8 radiation belts. Thus, the variations of the inner radiation belts are much longer than the time scales that are important for space weather events. The three types of motion that control the electron dynamics within the radiation belts are gyro motion, bounce motion, and drift motion.

Figure 3 demonstrates these motions via a schematic. Gyro motion refers to the motion of an electron under the influence of a magnetic field. The acceleration caused by the field results in the electron circling around magnetic field lines. The radius of the circle is referred to as the gyro radius and is larger for protons due to their mass and smaller for electrons. Bounce motion results from electrons undergoing changing magnetic field strength as they move from the magnetic equator to the magnetic pole. The final motion is the drift motion. Drift motion also arises from charged particles experiencing changing magnetic field strength. When particles encounter stronger fields, the increased force results in tighter gyro motion. The opposite is true for weaker fields. Therefore, the particles ‘drift’ perpendicular to the magnetic field vector in a direction based on their charge. Electrons drift eastward, while protons drift westward. Adding these three motions together can make for complicated electron pathways.

An important quantity in electron dynamics called the pitch angle becomes relevant with bounce motion. The pitch angle is the angle between the velocity vector of the electron and the magnetic field vector. As an electron enters a stronger magnetic field, such as when it approaches the magnetic pole, the pitch angle will increase until it reaches 90°. At this point, the electron reverses direction and returns back toward the opposite pole. The two locations (one for each hemisphere) where the electron’s pitch angle reaches 90° are called mirror points. The

‘bounce’ refers to the electron bouncing back and forth between these two points. When an electron has a small pitch angle at the magnetic equator, the mirror point will be further poleward

9 and the electron can penetrate deeper into Earth’s ionosphere before ‘bouncing’ back to its mirror point. The deeper the electron reaches in altitude, the more likely it is to ‘collide’ with a neutral molecule, creating an ion pair. An ion pair is both the electron that is freed from the collision and the positively charged atom/molecule that remains. When an electron bounces deep into the atmosphere and collides with the neutral molecules it is said to be lost or precipitated into the atmosphere. Therefore, as we will see, both internal and external forces that cause changes to the pitch angle distribution of a population of electrons can result in significant electron precipitation into the atmosphere.

To simplify the three motions, we can use conserved quantities called invariants that assume a slow, time-varying magnetic field. When the invariants are conserved, electrons are well behaved. However, when one or more of the invariants are violated, electrons can undergo acceleration that can change populations in the radiation belts and can lead to precipitation in the atmosphere. The first invariant describes the gyro motion. The second invariant describes the bounce motion. The third and last invariant refers to the drift motion. The three invariants mentioned above are described in further detail in Green (2010). Violations to the invariants occur when rapid energy changes to the electron occur on time scales associated with the invariants. Both internal and external forces can cause violations to the invariants. Examples of external forces include substorm acceleration, shocks from CMEs, CIRs, and radial diffusion.

Examples of internal forces include wave-particle interactions from chorus waves, electromagnetic ion cyclotron (EMIC) waves, and non-linear whistler waves (Green et al.,

2010). The three main processes believed to cause electron precipitation into the atmosphere are plasmaspheric hiss, EMIC waves, and very low frequency chorus waves (van de Kamp et al.,

2016; Summers et al., 2007; Millan and Thorne, 2007). EMIC waves are important for relativistic electron precipitation through near equator pitch-angle scattering (Rodger et al.,

2008; Meredith et al., 2011). Relativistic electrons are electrons that have velocities close to the speed of light and relativistic effects cannot be ignored. These electrons have energies above 1

MeV and can reach low altitudes in the atmosphere. EMIC waves, however, are only localized sources of precipitation. Plasmaspheric hiss is responsible for low levels of electron precipitation within the plasmapause. Chorus wave processes, on the other hand, are thought to be the dominant mechanism for electron precipitation during geomagnetic storms (Whittaker et al.,

2014).

The aurora is the visible manifestation of the last main source of energetic particles that influence Earth’s middle and upper atmosphere. Auroral electrons enter the atmosphere from the plasma sheet of the geomagnetic tail and in the polar-cusp region of the dayside magnetosphere

(Carleson Jr. and Egeland, 1995). Polar-cusp electrons have a narrow magnetic latitude range of a couple degrees centered around 78°, and precipitation is limited to the dayside magnetosphere.

They also have less energy than the plasma sheet electrons, with energies below 1 keV.

Nonetheless, they are visible optically at latitudes where the daytime sun is very low in the sky.

The more notable aurorae occur from continuous streams of low energy electrons from Earth’s plasma sheet that fall into the atmosphere into a region between 70o and 75o magnetic latitude called the auroral oval (Newnham et al., 2011). These electrons have energies between 1 keV and

10 keV and precipitate into the thermosphere above 100 km in the polar region (Eather et al.,

1976; Fang et al., 2008).

The last type of energetic particle sources that will be mentioned are galactic cosmic rays

(GCRs). GCRs can have enormous energies, but their influence on the middle and upper atmosphere is minimal because they usually deposit their momentum too low to influence the chemistry and dynamics of the middle atmosphere (Jackman et al., 1980). Thus, GCRs will not be considered in these investigations.

1.3. Particle Precipitation

Energetic particle precipitation (EPP) refers to the particles that enter the atmosphere from space. The precipitation of interest to this study is from SPEs, EEP events, and auroral electrons. Particle precipitation can be very important to middle and upper atmosphere chemistry during impulsive solar events due to the creation of odd nitrogen (N, NO, NO2) and odd hydrogen (H, OH, HO2) There are two categories of EPP effects on the atmosphere: the EPP- direct effect and the EPP-indirect effect (Randall et al., 2006; 2007). The direct effect refers to production of odd hydrogen (HOx) and odd nitrogen (NOx) by precipitating particles at the altitudes and locations where they deposit their energy. The EPP indirect effect refers to the

12 transport of NOx after production by EPP; because HOx has such a short lifetime in the middle atmosphere, it does not factor into the indirect effect. As mentioned in the previous section, photolysis is the primary loss mechanism of NOx, so in the polar winter NOx lifetimes in the mesosphere can be several months. This allows for transport of NOx from the mesosphere down to the stratosphere, where it can carry out catalytic ozone destruction. The following section reviews the nature of the various particle precipitation events and briefly reviews our current understanding of these processes.

SPEs can be defined as solar eruptions that result in protons from the sun precipitating into the atmosphere (Jackman et al., 2008). NOAA’s Space Weather Prediction Center defines a solar proton event as an integrated flux of > 10 MeV protons that exceeds 10 proton flux units, or an integrated flux of > 100 MeV protons that exceeds 1 proton flux unit at the primary GOES satellite (“GOES Proton Flux|NOAA” 2015). The proton flux unit is defined as the number of particles registered per second, per square cm, per steradian at the GOES detector. These types of events are usually caused by fast moving coronal mass ejections that erupt due to strong solar flares (Jackman et al., 2005). However, during high-speed solar wind streams from coronal holes, proton fluxes can increase as well and on rare occasion produce small SPEs. Solar protons penetrate the atmosphere to depths that depend on the energy of the proton. Figure 4 summarizes the range of altitudes over which the ionization occurs from various particles from EPP. It should be clear from the figure that solar protons penetrate the farthest due to their huge energies relative to photons and electrons.

The influence SPEs have on the atmosphere has been investigated thoroughly over the last few decades from both a modeling and observational perspective. Observational evidence of

SPE impacts on atmospheric chemistry has been explored by numerous authors and spans several instruments (Seppälä et al., 2007; Degenstein et al., 2005; Jackman et al., 1995; 2005;

2008; 2009; 2014; López-Puertas et al., 2005a; 2005b; Randall et al., 2001; Rohen et al., 2005;

Funke et al., 2011).

In late October and early November in 2003, major activity from the sun resulted in what is now commonly referred to as the ‘Halloween Storms’. The event began as a series of very strong solar flares that launched several CMEs, resulting in SPEs, and subsequent geomagnetic

14 activity. Thanks to observations during and around the event, the storms became a wealth of evidence of the consequences of impulsive solar activity. These observational studies not surprisingly show large increases in odd nitrogen between 50 and 90 km following the SPEs

(Seppälä et al., 2004; López-Puertas et al., 2005a). Decreases in ozone in the mesosphere and upper stratosphere can also been seen immediately following the two large SPEs in late October and early November (Degenstein et al., 2005; Jackman et al., 2005; Rohen et al., 2005). López-

Puertas et al. (2005a; 2005b) investigated several constituents during the Halloween storms using the Michelson Interferometer for Passive Atmospheric Sounding (MIPAS) instrument.

MIPAS was a high resolution, Fourier transform spectrometer that detected limb emission spectra between the wavelengths of 4.1 µm and 14.7 µm. High spectral resolution in the infrared spectral range enabled detection of many different atmospheric constituents, as described by

Funke et al. (2011). Results from the López-Puertas et al. (2005a) study showed large NOx enhancements over both polar caps, with less NOx enhancement occurring over the SH polar cap.

The durations of the enhancements were also asymmetric, with the NH showing longer duration

NOx enhancements than the SH. Decreases of 30-40% in NH ozone concentrations were seen days after the first SPE. HOx enhancements from the storms are inferred to be associated with immediate ozone depletion at the altitudes where the maximum ozone depletion took place (~55 km). Changes in ozone in the NH upper stratosphere of 30%, which occurred up to two weeks after the SPE (indirect effect), were associated with the observed NOx enhancements (López-

Puertas et al., 2005a). López-Puertas et al. (2005b) also looked at NOy (HNO3, N2O5, ClONO2) constituents during the Halloween storms. Enhancements in nitric acid (HNO3) were found both immediately after the SPEs and in late November through January in the upper stratosphere

(Orsolini et al., 2005). The first increase in nitric acid concentration is thought to be caused by

15 increases in OH and NOx. The second enhancement of nitric acid concentration is believed to be due to descent of NOx around the polar cap leading to the formation of N2O5. Through ion- cluster chemistry, N2O5 is converted to nitric acid. The pattern of two modes of enhancement in nitric acid is also found in N2O5 and ClONO2 concentrations following the storms (López-

Puertas et al., 2005b). These increases last until the end of December when the breakdown of the polar vortex forces the dispersion of NOy to lower latitudes.

The magnitude of the Halloween storms as well as the abundance of observations made this event popular for numerical modeling simulations. Funke et al. (2011) compared a number of different simulations of the Halloween storm SPEs to the MIPAS observations in order to evaluate model deficiencies during strong particle events. Many constituents were investigated, including the HOx, NOx, and ClOx families. In order to minimize differences between the models, the same ionization rates were used in each simulation. The ionization rates used were from the Atmospheric Ionization Module Osnabrück (AIMOS) model and included ionization from electron precipitation as well as proton precipitation (Wissing and Kallenrode, 2009). Not surprisingly, global climate models with nudged meteorology were found to perform better than models that were ‘free-running’. ‘Free-running’ refers to model simulations that do not use observations to nudge the dynamics of the model. Initial NOx enhancements and ozone depletion were shown to be fairly well represented by the model simulations when compared with MIPAS observations. However, some disagreements were found in odd nitrogen concentrations that descended into the stratosphere. This was thought to be due to dynamical differences between the models as well as lack of ion-ion and water cluster chemistry (Funke et al., 2011).

Particularly relevant to the work in this dissertation, MIPAS observations indicated increases in

16 odd nitrogen shortly after 20 November, but these increases were not simulated by any of the models, including WACCM. Funke et al. (2011) attributed this deficiency to incorrect simulation of the descent of odd nitrogen from above, and not to a lack of particle-induced production.

To summarize, previous studies on the modeling of SPEs show that chemistry changes simulated in the middle atmosphere compare reasonably well with observations. Both HOx and

NOx family production are well represented in the mesosphere and stratosphere during and after

SPEs. Mid-term (1 – 3 weeks) odd nitrogen enhancements in the stratosphere are also fairly well represented. Thus, ozone changes are well characterized shortly after proton events by models.

Most of the models, including WACCM, were not able to recreate the changes with minor nitrogen species such as ClONO2 and HNO3. This likely has to do with incomplete chemistry due to the lack of water cluster and ion-ion chemistry that had not been included in these models at the time the papers were written (Funke et al., 2011). WACCM has developed a version of

WACCM that includes D-region ion cluster chemistry that has since improved these results

(Verronen et al., 2016; Andersson et al., 2016; Orsolini et al. 2018).

The next type of EPP of interest is EEP. EEP events are strongest during times of enhanced geomagnetic activity when electrons in the radiation belts are lost to the bounce loss cone and precipitate into the atmosphere (Callis et al., 1991). However, auroral electrons are constantly precipitating into the upper atmosphere. In terms of chemistry changes in the middle and upper atmosphere, low energy electrons are an important source of thermospheric reactive

NOx. During times of strong geomagnetic activity, the flux of low energy electrons is thought to increase, creating more nitric oxide (NO) in the thermosphere (Solomon et al., 1999). These

17 increases would allow more NO to descend into the mesosphere under the right meteorological conditions. During solar maximum years, CMEs are often the cause of geomagnetic storms that change the electron flux in the radiation belts. However, during the declining stage of solar maximum and during solar minimum, co-rotating interaction regions (CIRs) are usually the driver of geomagnetic storms, through changes in the solar wind speed (Rodger et al., 2010).

Electron populations in the radiation belts are also influenced by the 27-day solar rotation, regardless of the solar cycle (Callis et al., 1991). Some of the trouble with attempting to quantify electron precipitation is that some electrons are lost out of the belts and precipitate into the atmosphere, while some are lost to the magnetopause (Turner et al., 2012b). Recently, much discussion has taken place regarding the drivers of the rapid changes in radiation belt electron flux and the factors that determine whether the electrons precipitate into the atmosphere. It is currently thought that pitch-angle scattering through EMIC wave-particle interactions is the primary mechanism that allows relativistic electrons to precipitate into the atmosphere during the recovery phase of the storm (Turner et al., 2014). Relativistic electrons that are lost during the main phase of the geomagnetic storm are thought to undergo outward radial transport

(Jordanova et al., 2008). As previously mentioned, chorus waves and plasmaspheric hiss are thought to be the primary drivers of energetic electron precipitation.

The electrons that do precipitate in the atmosphere penetrate to depths based on their energy, as summarized in Figure 3. The influence electrons have on the atmosphere is similar to protons in that they create odd hydrogen and odd nitrogen through the creation of ion pairs

(Solomon et al. 1981). However, since electrons have smaller energies than protons, they tend to deposit the majority of their momentum higher in the atmosphere than protons. Relativistic

18 electrons are the exception to this rule. These high-energy electrons (> 1 MeV) have speeds close to that of light and can reach depths of 50-70 km (Thorne et al. 1980; Callis et al. 1991; Frahm et al. 1997). Another aspect of relativistic electrons is that they may lead to an additional source of ionization. When relativistic electrons precipitate into the atmosphere, they emit

Bremsstrahlung x-rays. The radiation occurs due to relativistic electrons being rapidly decelerated as they enter the denser atmosphere. Although most of this radiation is thought to be emitted at 80-100 km, it can propagate into the stratosphere. Therefore, these x-rays could add an additional ionization source around 35-40 km (Frahm et al., 1997). However, recent work shows that Bremsstrahlung radiation is not likely a significant source of ionization in the atmosphere

(Xu et al., 2018).

Some studies have looked at the effects of relativistic electron precipitation (REP) events.

Using data from three Charged Particle Analyzer (CPA) instruments and three Spectrometer for

Energetic Electrons (SEE) instruments aboard three different spacecraft, Callis et al. (1991) investigated the effects of relativistic electrons and Bremsstrahlung radiation on odd nitrogen and odd hydrogen in the middle atmosphere. The CPA and SEE instruments both rely on electron counting in different energy bins. The CPA instruments include 12 energy levels from 30 keV to

2 MeV. The SEE instruments have four energy bins from 3 to 15 MeV (Callis et al., 1991). Both short-term and long-term ozone destruction was found by including relativistic electron counts from the instruments into a 2D photochemical model. Comparisons were made for NO2 with

LIMS, SAGE I, and SAGE II data. The comparisons showed improvement when adding electrons into the model versus simulations without energetic electrons. A more recent study looked at relativistic electron precipitation from microburst events (Seppälä et al., 2018).

Previously, microbursts were not thought to influence mesospheric chemistry (Turunen et al.,

2009). Nonetheless, Seppälä et al. (2018) found increases in HOx and NOx that resulted in ozone losses of 10 – 20% in the mesosphere. Thus, including relativistic electrons in model simulations would likely improve the accuracy of middle atmosphere constituents during and after geomagnetic activity.

Medium energy electrons (MEEs) (30 keV – 1 MeV) have been understudied relative to other forms of particle precipitation in the atmosphere. The difficulty has to do with our limited ability to measure these electrons accurately. Codrescu et al. (1997) studied the influence of

MEEs on the mesosphere and thermosphere using the National Center for Atmospheric Research

(NCAR) Thermosphere-Ionosphere-Mesosphere Electrodynamics General Circulation Model

(TIME-GCM) and MEPED data. The results showed that MEEs do influence NOx and HOx concentrations, leading to ozone depletion in the mesosphere. However, the electron data used in the study was found to be contaminated by protons (Lam et al., 2010; Rodger et al., 2010; Peck et al., 2015). Because of this limitation, global climate models like WACCM have usually excluded MEEs as a particle source. Therefore, these simulations underestimate the amount of

EEP-induced ionization in the atmosphere. Without the proper ionization profiles, chemistry changes in the middle atmosphere are incomplete and underestimated.

Recently, more work has been done to investigate the impact of energetic electrons. This is partly due to advancements in processing the MEPED data set. Several MEE data sets now exist that use various methods to account for proton contamination and to compute differential electron and proton spectra from the coarse spectral bins of the MEPED instrument (Wissing and

Kallenrode 2009; Lam et al., 2010; Asikainen and Mursula 2013; Green et al., 2013; Peck et al.,

2015; Nesse-Tssøy et al., 2016; van de Kamp et al., 2016). A new development from Yando et al.

(2012) improved the accuracy of the geomagnetic factors that are used to convert the MEPED count rates into differential fluxes. These advances paved the way for much improved electron data sets. Studies have attempted to evaluate POES MEPED observations using riometers and ground-based observations (Clilverd et al., 2013; Rodger et al., 2013). These results showed that

POES is likely underestimating flux rates due to diffusion causing pitch-angle scattering into the bounce loss cone. Nonetheless, improvements are being made to further increase the accuracy for global climate model studies using electron data sets.

For the Coupled Climate Model Intercomparison Project version 6 (CMIP6), the

Stratospheric Processes and their Role in Climate group organized the creation of an electron precipitation data set (van de Kamp et al., 2016; Matthes et al., 2017). Andersson et al. (2018) utilized the CMIP6 EEP data set to investigate polar ozone losses from MEE precipitation. They found MEE to have a significant impact on ozone losses in both the mesosphere and the stratosphere. Other recent studies have taken a parameterized approach, using geomagnetic indices as proxies for electron precipitation. Hendrickx et al. (2017) investigated electron precipitation in the atmosphere caused by the 27-day solar rotation cycle. They used two geomagnetic indices (AE, Ap) rather than POES MEPED data, and still found significant NO increases in the mesosphere. Another recent study used simulations from WACCM and

WACCM-D (D-region chemistry) both with and without MEEs to attempt to recreate the April

2010 EEP event (Smith-Johnsen et al., 2018). They found that using D-region chemistry with

WACCM improved the direct production of NO versus WACCM without D-region chemistry.

Strong correlations between the model and observations were found in the lower thermosphere region and in the middle mesosphere; however, near the mesopause the model appears to underestimate NO. They concluded that the model dynamics are still lacking, resulting in insufficient descent across the mesopause.

1.4. Middle Atmosphere Chemistry

Two drivers of chemistry need to be considered when studying space weather effects on the middle atmosphere. The first is energetic photons that dissociate neutral molecules through photoionization on the dayside of the Earth. The second is energetic charged particles (protons, electrons), which follow magnetic field lines and therefore deposit most of their energy in polar latitudes. Photons impart all of their energy during the initial impact, which creates photon electrons that take much of the original photon energy with them. Charged particles continue to penetrate into the atmosphere after initial collisions, creating more ion pairs until depositing all of their energy. As we will demonstrate, both photons and charged particles create ion pairs that can change middle atmospheric chemistry through the creation of odd nitrogen and odd hydrogen.

Photons of all wavelengths are constantly bombarding the Earth and its atmosphere.

Since the sun’s effective temperature in the photosphere is approximately 5700 K, according to

Planck’s law, the sun’s emission peak is in the visible part of the electromagnetic spectrum.

During solar flares, increases occur primarily in the radio, x-ray and UV part of the spectrum

(Dennis, 1988). Photons of the two latter wavelengths can penetrate to the lower thermosphere and mesosphere. Figure 5, from Solomon et al. (2005), shows the altitudes where photons of x-

22 ray and UV wavelengths deposit their energy. For the thermosphere, x-rays and EUV wavelengths are important, whereas for the mesosphere and stratosphere specific bands such as

Lyman-alpha and longer UV wavelengths are important. Nonetheless, much longer wavelengths like radio waves can create radio bursts during solar flares, which can disrupt aviation and

Global Positioning System (GPS) communications (Cane et al., 2002).

Radiation of all wavelengths entering the atmosphere follows Beers law:

퐼(푧) = 퐼∞ exp (−휏(푧)) (1.1)

Here I is the radiation intensity at a given altitude ‘z’, τ is the optical depth and I∞ is the radiation intensity at the top of the atmosphere. What this equation shows is that the penetration depth of radiation of a particular wavelength is dependent on the optical depth, which is dependent on solar zenith angle, column density, and the absorption cross section, as shown below:

휎푁(푧) 휏(푧) = (1.2) cos (휃) Here θ is the solar zenith angle, N is the column density and σ(λ) is the absorption cross section.

The energy deposition caused by a particular wavelength of radiation (λ) is given by:

푞휆(푧) = 퐼휆(푧)휎푁(푧) (1.3)

Therefore, the ionization at a given altitude, qλ(z), is equal to the spectral radiation multiplied by the cross section of the potential collisions and the number of possible targets. By utilizing the scale height, H, and assuming an exponential decay of neutral molecules with height and exponential decay of radiation with depth, we can derive Chapman’s function:

푧−푧 푞(푧) = 퐼 휎푛 exp(− 0 − τ(z)) (1.4) ∞ 0 퐻

Where 푧0 is some reference altitude and 푛0 is the column density at that altitude. Figure 4 shows the expected energy deposition altitude of photons as a function of wavelength using the

Chapman function. We can see that short wavelengths (< 10 nm) can ionize the mesosphere, while the longer wavelengths (> 10 nm) are mostly influential in the thermosphere. For a more detailed review of photoionization and Chapman’s function, the reader is encouraged to see

Fuller-Rowell and Solomon (2010) and Luhmann (1995).

When energetic charged particles enter the polar atmosphere along magnetic field lines, they eventually encounter the neutral atmosphere. Collisions with neutral atmospheric constituents increase as air density increases. These collisions cause ionization, excitation, dissociation, and dissociative ionization of atmospheric molecules, which produce showers of electrons and ions. Additional electrons freed from collisions in particle showers dramatically increase the production of chemically active HOx and NOx (Solomon et al., 1981; Thorne et al.,

1980).

In the upper stratosphere and mesosphere, the most important constituents for particle collisions are N2 and O2; atomic oxygen is also important in the thermosphere above 100 km, as it exceeds the molecular oxygen mixing ratio here (Goody and Yung, 1989). Secondary electrons are electrons that are removed from neutral constituents. Once the secondary electrons are detached, the molecule or atom that is left becomes a positive ion. Secondary electrons lose kinetic energy during the collisions with other neutral molecules, which produces more ions and secondary electrons. The species most efficiently produced in the middle atmosphere from ion pair production are H, OH, NO, and N. As the chemistry will show, this is due to the abundance of oxygen and nitrogen in the atmosphere.

Regardless of process, both photo-ionization and particle precipitation create ion pairs and secondary electrons. Beginning with primary and secondary electrons (e*), the following reactions describe possible pathways for odd hydrogen production. First, the secondary electrons collide with oxygen molecules, creating additional electrons following one of two pathways:

Secondary electron collisions

∗ + 푒 + 푂2 → 푂2 + 2푒 (R 1.1)

∗ + 푒 + 푂2 → 푂 + 푂 + 2푒 (R 1.2)

The positive oxygen ion can then react with water vapor to produce a hydroxyl radical (OH), which is one of the most reactive constituents in the atmosphere because of its unpaired electron.

The three primary pathways for OH production are:

Odd Hydrogen Pathway (1):

+ + 푂2 + 푂2 + 푀 → 푂4 + 푀

+ + 푂4 + 퐻2푂 → 푂2 ∙ 퐻2푂 + 푂2

+ + 푂2 ∙ 퐻2푂 + 퐻2푂 → 퐻3푂 ∙ 푂퐻 + 푂2

+ − 퐻3푂 ∙ 푂퐻 + 푒 → 퐻 + 푂퐻 + 퐻2푂

Net: 퐻2푂 → 퐻 + 푂퐻 (R 2.1)

Odd Hydrogen Pathway (2):

+ + 푂2 ∙ 퐻2푂 + 퐻2푂 → 퐻3푂 ∙ 푂퐻 + 푂2

+ + 퐻3푂 ∙ 푂퐻 + 퐻2푂 → 퐻3푂 ∙ 퐻2푂 + 푂퐻

+ + 퐻3푂 ∙ 퐻2푂 + 푛퐻2푂 → 퐻3푂 (퐻2푂)푛+1

+ − 퐻3푂 (퐻2푂)푛+1 + 푒 → 퐻 + (푛 + 2) 퐻2푂

Net: 퐻2푂 → 퐻 + 푂퐻 (R 2.2)

Odd Hydrogen Pathway (3):

+ + 푂2 ∙ 퐻2푂 + 퐻2푂 → 퐻 ∙ 퐻2푂 + 푂퐻 + 푂2

+ + 퐻 ∙ 퐻2푂 + 푛(퐻2푂) → 퐻3푂 (퐻2푂)푛

+ − 퐻3푂 (퐻2푂)푛 + 푒 → 퐻 + 푛(퐻2푂)

Net: 퐻2푂 → 퐻 + 푂퐻 (R 2.3)

Note that the availability of water vapor is critical to the formation of H and OH in the three odd hydrogen pathways, so these reactions are only important for altitudes where water molecules are available. Water vapor in the middle atmosphere decreases significantly with height resulting in

26 limited water above the mesopause. Odd hydrogen formed this way can go on to destroy ozone via three main catalytic reactions that depend on altitude. The first reaction below is important above 40 km, where energy deposition from energetic particles is prevalent during precipitation events. Likely, this dominant reaction destroys ozone in the mesosphere immediately following an EPP event:

퐻 + 푂3 → 푂퐻 + 푂2 푂퐻 + 푂 → 퐻 + 푂2

Net: 푂3 + 푂 → 2 푂2 (R 2.4)

The competing reaction that destroys ozone in the upper stratosphere is:

푂퐻 + 푂3 → 퐻 + 2 푂2 퐻 + 푂2 + 푀 → 퐻푂2 + 푀 퐻푂2 + 푂 → 푂퐻 + 푂2

Net: 푂 + 푂 + 푀 → 푂2 + 푀 (R 2.5)

The two other reactions that include odd hydrogen are only important to the ozone budget below

40 km. These are:

푂퐻 + 푂3 → 퐻푂2 + 푂2 퐻푂2 + 푂 → 푂퐻 + 푂2

Net: 푂3 + 푂 → 2 푂2 (R 2.6)

And

푂퐻 + 푂3 → 퐻푂2 + 푂2 퐻푂2 + 푂 → 푂퐻 + 푂2

Net: 푂3 + 푂 → 2 푂2 (R 2.7)

The other species created from collisions with ion pairs is NOx. The main source of NOy in the stratosphere is from reactions with O(1D), which is created through photolysis of ozone,

27 and N2O that has been transported from the troposphere (Thorne et al., 1980). This reaction is:

1 푁2푂 + 푂( 퐷) → 2 푁푂 (R 3.1)

The creation of O(1D) requires sunlight, which is strongest at the equator. When no sunlight is present, NO is converted to NO2 via the following reaction:

푁푂 + 푂3 → 푁푂2 + 푂2 (R 3.2)

During EPP events, odd nitrogen is created through collision chemistry in a manner similar to the production of odd hydrogen. Currently, it is assumed that 1.25 nitrogen atoms are produced per ion pair (Porter et al., 1976). The primary reactions pertinent for odd nitrogen production from secondary electrons are discussed in detail in Rusch et al. (1980) and will be reviewed briefly here.

Secondary electrons have three production pathways when colliding with a neutral nitrogen molecule. These are:

∗ + 푒 + 푁2 → 푁2 + 2푒 (R 3.3)

∗ 4 2 2 + 푒 + 푁2 → 푁( 푆, 퐷, 푃) + 푁 + 2푒 (R 3.4)

∗ 4 2 2 푒 + 푁2 → 푁( 푆) + 푁( 퐷, 푃) + 푒 (R 3.5)

The first of these reactions frees an electron to add to the secondary electron as well as a positive nitrogen ion. The nitrogen ion has three possible pathways (Rusch et al., 1980); however, the dominant pathway is the ion reacting with another electron to produce two nitrogen atoms.

∗ + 2 4 푒 + 푁2 → 푁( 퐷) + 푁( 푆) (R 3.6)

This means that all of the reactions that occur between neutral nitrogen molecules and electrons

28 result in atomic nitrogen in one of three states. These are the ground states N(4S) and N(2P) as well as the excited state N(2D), which plays the most important role of creating NO during an

2 EPP event. This is because N( D) immediately reacts with O2 to create NO, following:

2 푁( 퐷) + 푂2 → 푁푂 + 푂 (R 3.7)

The ground state of nitrogen N(4S), on the other hand, actually destroys NO through the following cannibalistic reactions:

4 푁( 푆) + 푁푂 → 푁2 + 푂 (R 3.8)

4 푁( 푆) + 푁푂2 → 푁2푂 + 푂 (R 3.9)

Thus, the ratio of formation between ground state nitrogen and excited state nitrogen dictates NO production. From the ~1.25 N atoms created for each ion pair, the suggested branching ratio is

0.7 N(2D) atoms and 0.55 N(2S) (Porter et al., 1976; Jackman et al., 2009). During EPP events, odd hydrogen (specifically OH) is the dominant ozone destroyer in the mesosphere. Odd nitrogen also plays a role in ozone destruction immediately following the event. However, as explained later, in the polar winter region odd nitrogen can also be important several weeks or months after the event. The chemical reaction that is important for ozone destruction via NOx is:

푁푂 + 푂3 → 푁푂2 + 푂2 푁푂2 + 푂 → 푁푂 + 푂2

푂3 + 푂 → 2 푂2 (R 3.10)

The reactions shown above clearly demonstrate how EPP can create the reactive species

HOx and NOx, which results in the systematic destruction of ozone. To reiterate, in the mesosphere, HOx is most important immediately following its creation and has a very short residence time due to its reactive nature. NOx, on the other hand, can have long residence times

29 in the middle atmosphere when there is no sunlight to destroy it. While EPP-induced production of NOx occurs primarily in the thermosphere and mesosphere, descent in the polar winter can bring reactive NOx into the stratosphere where it is most effective at destroying ozone. The lifetime NO can be seen in Figure 6. The blue line shows model simulation results from

WACCM during the SH winter, while the red line shows NO lifetimes in the SH summer.

During the winter, in polar latitudes where sunlight is limited, NO can have very long lifetimes and can be controlled dynamically. This is evident looking at the lower mesosphere and stratosphere in Figure 6a, which show lifetimes of well over 100 days. During the summer, this is much shorter resulting in limited NO transport in this region of the atmosphere. NO’s seasonal lifetime differences differs from other chemical changes caused by energetic particle precipitation such as OH and is one of the primary reasons the NOx family in the middle atmosphere is so important during impulsive solar events.

1.5 Meteorology

Meteorology plays a significant role in the chemistry of the middle atmosphere. NOx transported from the MLT to the stratosphere is capable of destroying ozone (e.g., Solomon et al., 1982; Randall et al., 1998) or otherwise interfering with ozone catalytic cycles (Jackman et al., 2008). As reviewed above, HOx is very reactive, but has a short lifetime in the middle atmosphere and therefore primarily affects the chemistry by means of the direct effect. NOx differs in that its lifetime depends on the amount of solar irradiance available. In the polar winter when little to no light is available, NOx can be controlled dynamically and descend into the stratosphere where it can influence local chemistry through the indirect effect.

The balance between radiative cooling and ultraviolet heating from ozone sets up the circulation of the middle atmosphere (Andrews 1987; Holton 2004). Unlike the troposphere, which has a surface to radiate excess heat, in the stratosphere ozone maintains thermal equilibrium. The zonal mean circulation of the middle atmosphere is represented in Figure 7.

Both the winter and summer hemispheres have strong wind patterns called jets, similar to the jet streams we find in the troposphere. The polar night jet in the winter hemisphere is closer to the pole and stronger than the summer jet. During the polar winter, the strong latitudinal gradient in temperature between the equator and pole creates an intense circulation through thermal wind balance. The circulation that separates the dark, wintertime polar regions and the mid-latitudes in the middle atmosphere is often referred to as the polar vortex (Andrews et al., 1987; Nash et al.,

1996; Harvey et al., 2015). The polar vortex is not limited to the stratosphere, but reaches up through the mesosphere to the mesopause (Harvey et al., 2015). In the middle atmosphere, it acts as a containment vessel for minor atmospheric constituents like odd nitrogen (Manney et al.,

2000). Inside the polar vortex, cold air descends, bringing minor constituents such as NOx

31 downward. During years of an undisturbed polar vortex, air from the mesosphere has weeks to months to descend into the stratosphere. This is common in the SH where little disruption occurs to the polar vortex. In the NH, this is not the case, as explained next.

Variations in the middle atmosphere circulation during the polar winter are strongly linked to the propagation and breaking of planetary waves. Disruptions to the polar vortex can occur often from the propagation of wave number 1 and 2 planetary waves from the troposphere

(Siskind et al., 2010). These waves are created by land-sea temperature contrasts as well as large- scale topography. Figure 8 shows the general meridional circulation as well as the location of planetary and gravity wave activity in the middle atmosphere. Under some conditions, planetary wave breaking in the stratosphere can induce sudden stratospheric warming events (SSWs)

(Andrews et al., 1987; Manney et al., 2008; Manney et al., 2009a; Siskind et al., 2010). Enough wave breaking leads to a halt or even reversal of the stratospheric mean flow. Since the SH is

32 mostly ocean, it sees less planetary wave flux than the NH. The NH, on the other hand, sees significant planetary wave activity, resulting in a more variable middle atmosphere during the polar winter and consequently, more variable descent rates. Without the influence of planetary waves there is little disruption to the polar vortex and the descent rates in the SH are less variable

(e.g. Siskind et al., 2007; Randall et al., 2007).

Changes in the circulation in the stratosphere from SSWs will change the gravity wave filtering, thereby influencing descent rates in the mesosphere (Manney et al., 2009a; Manney et al., 2009b; Thurairajah et al., 2010; Holt et al., 2013). Strong and prolonged SSWs can be associated with elevated stratopause (ES) events, which occur during the recovery phase of the

SSW. During an ES event, wave-induced enhancement of the downward branch of the residual

33 circulation results in reformation of the stratopause at a much higher altitude than usual (Manney et al., 2010; Siskind et al., 2010). Research has shown that ES events can have large impacts on descent rates in the winter polar region (Funke et al., 2005; Siskind et al., 2007; Randall et al.,

2009). Therefore, if EPP events occur shortly after an ES event occurred, this can result in large

NOx concentrations descending into the stratosphere and destroying ozone. The Arctic winter of

2004 was one such time period, when enhanced geomagnetic activity was coupled with an ES event during the NH polar winter (Randall et al., 2005; Randall et al., 2015). This particular winter was an anomalous winter relative to other NH winters on record (Manney et al., 2005; Jin et al., 2005; Randall et al., 2015). In January, a SSW caused an extended ES event that resulted in strong decent rates in the NH mesosphere and upper stratosphere. In addition to the anomalous dynamics, strong geomagnetic activity occurred throughout the first two months of 2004, increasing the electron precipitation in the mesosphere and lower thermosphere. This resulted in record levels of NOx descending from the mesosphere into the stratosphere in the spring months

(Randall et al., 2005). Models are currently unable to simulate accurately this time period, likely due to a combination of poorly resolved mesospheric dynamics as well as insufficient ionization without including medium and high-energy electrons in the models.

1.6.1 Whole Atmosphere Community Climate Model

The Whole Atmosphere Community Climate Model (WACCM) is a superset of the

Community Atmosphere Model (CAM) run in the numerical framework of the Community Earth

System Model (CESM); it is a fully coupled global chemistry climate model. The version of the model used in the studies presented here is WACCM4, which is based on CAM version 4.

Details about the WACCM model can be found in Chapter 2 and Chapter 3. The module for

34 atmospheric chemistry is from the Model for Ozone and Related Chemistry Tracers (MOZART) version 3, which is described in detail in Kinnison et al. (2007). There are 59 species with 217 gas-phase reactions as well as 17 heterogeneous reactions.

One of the best way to constrain dynamics in WACCM is to use the configuration set called specified dynamics WACCM (SD-WACCM). SD-WACCM utilizes data from NASA’s

Modern-Era Retrospective Analysis for Research and Applications (MERRA) reanalysis model to nudge the dynamics of WACCM each time step (Lamarque et al., 2012; Rienecker et al.,

2011). This is performed by combining 1% of the reanalysis data and 99% of the model- calculated data for dynamical fields in the model every 30 minutes. The nudging is linearly reduced from 40-50 km, where the model becomes free running i.e., not nudged. Using SD-

WACCM also increases the number of pressure levels from 66 to 88. When investigating chemical changes from energetic particles, constraining the dynamics in as much of the atmosphere as possible is the best way limit the variability of dynamics between simulations.

The studies described in this dissertation used three different versions of WACCM. The first study used WACCM4, but with a reduced time step from 30 minutes to 5 minutes. We needed to reduce the model time step for the inclusion of a high time resolution photon data set.

Solar flares only last several minutes to an hour. In order to capture the effects of the various phases of the flare, the highest temporal resolution possible is needed. The photon data set will be explained in detail below. The second study also used WACCM4 but included D-region ion chemistry. The D-region chemistry module adds 307 reactions in addition to the default

MOZART chemistry (Verronen et al., 2016). Many of these reactions are from water cluster ions that are common in the mesosphere. As shown by Andersson et al., (2018) and Smith-Johnsen et al. (2018), WACCM-D more accurately describes the mesosphere and lower thermosphere

35 chemistry during geomagnetically active times. Since energetic electrons will deposit most of their energy in the MLT region, it is important to have the most accurate chemistry available.

1.6.2 Ionization from Protons

For the work in this dissertation, it was critical to include EPP-induced ionization in the various versions of the WACCM model that were employed. Computation of ionization rates from solar protons were based on the method of Vitt and Jackman (1996). Using this method, proton ionization profiles are calculated using data from the Interplanetary Medium Platform instruments 1-8 from 1963 – 1993, and the Geostationary Operational Environmental Satellites

(GOES) since 1993. The proton fluxes are provided using 6 energy intervals from > 1 MeV to >

100 MeV every 5 minutes. Once the proton fluxes are compiled, a power law spectrum is fit to the proton channels in order to derive the differential proton energy spectrum, which is then split into 60 monoenergetic intervals. The estimation from Porter et al. (1976) is that ~35 eV is needed to create one ion pair. Thus, once the energy profiles are created, the ion pair profiles are computed by dividing the fluxes by 35 eV.

The profiles built from this methodology are typically made into daily averages.

However, to increase the accuracy of the studies presented here, we used hourly averages. The energy profiles are averaged over the polar caps from 60° - 90° magnetic latitude, with a vertical range of 888 hPa to 8 x 10-5 hPa. Since WACCM is on a geographic, hybrid sigma pressure coordinate system, the magnetic latitude is converted to the WACCM latitude-longitude grid and the vertical profiles are converted to the WACCM pressure grid.

Unlike in WACCM-D, WACCM does not have ion pair chemistry built into the model in the mesosphere. Instead, once the ion pair profiles are created, each ion pair is assumed to create

HOx and NOx molecules based on parameterizations. For odd hydrogen, constituents are produced using a table given in Jackman et al. (2005) and shown below in table 1. At low altitudes (40 – 60 km), HOx production is approximately two molecules per ion pair. The production decreases with increasing altitude and increasing ionization rate. For odd nitrogen,

1.25 nitrogen atoms are assumed for each ion pair. The atoms are split up between the ground state, N(4S), and the excited state, N(2D), using a branching ratio of 0.55 per ion pair for the ground state and 0.7 per ion pair for the excited state, respectively.

1.6.3 Ionization from Electrons

Two types of electron precipitation were incorporated into the WACCM studies described below. The first is the auroral ionization. The aurora in WACCM is currently parameterized using methodology described in Roble and Ridley (1987). The process estimates polar cap energy using geomagnetic indices to produce energy deposition profiles that are polar

37 cap averaged onto the WACCM grid. This methodology has been used successfully for the aurora in CESM for many years.

The second type of electron precipitation is medium (30 keV – 1 MeV) and high-energy

(> 1 MeV) electron precipitation. As mentioned in the previous section, several MEE data sets exist, all of which use slightly different methods. Both of the MEE data sets used in this study utilize the POES MEPED. The MEPED detectors used in this study are a part of the Space

Environment Monitor -2 (SEM-2) instruments that fly aboard the Polar Operational

Environmental Satellites (POES) and the MetOp-2 satellites. The satellites have a near circular orbit at about 850 km. Each MEPED instrument consists of two proton telescopes and two electron telescopes. Each telescope pair has a telescope that points toward the solar zenith (0° detector), while the other points away from the antiram direction (90° detector). Each telescope has a 30° field of view with a combined 2-second sampling time using both detectors. The proton telescopes have 6 differential energy channels with an energy range of 30 keV – 6900 keV labeled P1 – P6. The proton telescopes are designed with cobalt magnets that bring incident electrons into an aluminum bin, preventing them from reaching the detector. The electron telescopes have three integrated energy channels including > 30 keV, > 100 keV, and > 300 keV channels. The electron telescopes have a nickel-foil cover to prevent low energy protons from entering the detector. The electrons that reach the detector collide into a silicon sheet that can absorb electrons with energies up to 2500 keV. For a full description of the MEPED instruments, the reader is encouraged to consult Evans and Greer (2000).

When using the MEPED data set, the electron count rates must be corrected for proton contamination. One of the popular methods for removing the protons from the electron channels is to fit a mathematical spectrum to the proton channels, creating a differential flux. The spectral

38 fit that is most often assumed is a power law spectrum. Once a differential proton flux is created, specific proton energy ranges can be isolated and subtracted from electron channels where contamination is suspected. Once the electron channels are free from proton contamination, they can be used to calculate electron fluxes, which can be transformed into ionization rate profiles for global climate models.

The first electron data set used in this dissertation is from the Coupled Climate Model

Intercomparison Project version 6 (CMIP6) (Matthes et al., 2017). Many forcing data sets were created in an effort to support the intercomparison project. One such data set was an energetic electron data set used for estimating effects from geomagnetic storms. The data set is explained in detail in van de Kamp et al. (2016) and again in Andersson et al. (2018). The CMIP6 electron data set is based on measurements from the 0° MEPED detectors. The method described in Lam et al. (2010) is applied in order to remove proton contamination from the electron measurements.

The corrected electron counts are converted to directional fluxes using the conversion method described in Evans and Greer (2004).

After calculating the spectral flux for the time period from 2002-2012, the data are used to derive a relationship between the fluxes and the Ap index for that time period. This relationship is then used to infer the precipitating spectral flux for any given value of the Ap index, and these Ap-based fluxes are then used to compute vertical profiles of ionization rates via the mono-energetic method in Fang et al. (2010). The ionization rate calculations use temperatures and densities from an atmospheric model described in Picone et al. (2012). The resulting data set contains daily ionization rates binned by L-shell. An additional step was taken in this study to interpolate the CMIP6 ionization rates to the WACCM grid by converting L-shell to magnetic latitude and finally to geographic latitude.

The second electron data set used here is based on Peck et al. (2015), which will be referred to as P15 from hereon. Similar to the CMIP6 MEE data set, the method of P15 constructs a differential electron flux spectrum from the MEPED count data. However, there are differences between the two data sets. The first is that P15 does not assume a specific functional fit to describe the spectral dependence of the measured proton or electron count rates. Rather,

P15 generates a linear combination of four functions and uses least square regression to calculate the most accurate combination. The four functions assumed are the power law, energy exponential, Maxwellian, and the double Maxwellian. We believe this methodology is more accurate than assuming a specific function such as a power law to create the differential spectrum. Once the functional form is fit to the spectrum, the same method of removing contaminated protons from the electron channels is employed.

Another difference is P15 uses a 5-day integration method using all available MEPED instruments in order to get a 3D energy flux map. In contrast, CMIP6 bins fluxes by l-shell, which can be converted to magnetic latitude. Thus, CMIP6 is zonally averaging by magnetic latitude, rather than calculating a complete 4D picture. Once the energy flux maps are created, they are used as inputs into WACCM. P15 follows the methods of Fang et al. (2010) to compute ionization rate profiles as well; however instead of assuming a standard temperature profile for the densities, P15 uses the temperature profile calculated from WACCM.

The methods of the Fang et al. (2010) parameterization to compute ionization rates from the CMIP6 and P15 data sets will be described here. Previously, range calculations were used in order to parameterize the impact of energetic electron precipitation (Spencer et al. 1959; Lazarev et al. 1967; Lummerzheim et al. 1992). These methods were limiting because they were primarily scaled laboratory measurements that did not predict atmospheric collisions correctly.

To circumvent this, Fang et al. (2008; 2010) used two detailed first-principles models to calculate Boltzmann transports equations. The results from these models are then normalized before curve fitting is performed. The range of the model is monoenergetic bins between 100 eV and 1 MeV. In the case of the P15 data set, 23 separate energy bins are used. As mentioned above, a temperature profile is needed to compute background densities for the ionization rate calculations. CMIP6 used atmospheric temperatures and densities from Picone et al. (2012) and the rates are computed offline. They are then averaged on l-shell and are mapped to the

WACCM grid in order to be used in simulations. P15 uses the WACCM model temperatures and densities within the simulation. Thus, the rates are computed online while the simulation is running. While it would appear that this could make a large impact on the results, Fang et al.

(2008; 2010) showed that this should not make a large difference on the ionization rate calculations.

Some modifications were made to the data set from P15 to produce a new, modified P15

(MP15) data set. One modification was the limitation of l-shells at low and high geomagnetic latitudes. With l-shells below 2, the POES MEPED instrument becomes unreliable. This is due to the fact that the 0° detector at low magnetic latitudes no longer views the bounce loss cone

(BLC), and instead is more likely viewing trapped flux and/or the drift loss cone. Thus, we cannot discern what fluxes are truly precipitating and what fluxes are trapped. The higher l-shells are excluded to match previous work done using the MEPED data set (Rodger et al., 2010; van de Kamp et al., 2016). The second modification made to P15 was adding a noise floor. The

MEPED instrument has a minimum detection level of 250 electrons cm-2 sr-1 s-1 in the lowest energy channel. If the count rate is below this level, we cannot distinguish between counts and noise. Therefore, during times when the E1 channel count rates are below this threshold, all bins

41 are zeroed. The addition of the noise floor makes a particularly large impact on the P15 data set because of the interpolation method it employs.

1.6F.4 Ionization from Photons

Solar flares enhance ionization in the mesosphere and thermosphere via high-energy photons. To include flare ionization in WACCM and WACCM-X, versions of the CESM model use the Flare Irradiance Spectral Model (FISM). FISM is an empirical model that estimates irradiance in 1 nm wide bins from 0.1 nm to 190 nm with 1-minute time resolution. For a complete description of the model, readers are referred to Chamberlin et al. (2007; 2008). The irradiances are calculated using six proxies from instruments aboard the Thermosphere

Ionosphere Mesosphere Energetics and Dynamics (TIMED) satellite, as well as the Upper

Atmosphere Research Satellite (UARS) and Solar Radiation and Climate Experiment (SORCE) satellite. For short wavelength estimates (0.1 – 10 nm), the TIMED Solar EUV Experiment

(SEE) XUV Photometer System (XPS) observations are used along with SORCE observations.

For the EUV (10 nm – 120 nm), the TIMED SEE observations are used. For the longer wavelengths (121 nm – 200 nm), SEE is used along with UARS Solar Stellar Irradiance

Comparison Experiment (SOLSTICE). Since the model is empirical, the flare components are initialized using a database of 39 large flares from the TIMED SEE instrument. The proxies used to estimate the irradiances include F10.7 and Mg II, as well as Ly-α, the 0 – 4 nm integrated band, the 1 nm band centered at 36.5 nm, and the 1 nm band centered at 30.5 nm, all from

TIMED SEE.

When computing daily irradiances, the model uses baseline values from observations and estimates effects from solar rotation and the solar cycle that are then added to the baseline.

Equation 2.1 shows the calculation in mathematical form:

퐸(휆, 푡) = 퐸푚푖푛 + Δ퐸푆퐶(휆, 푡) + Δ퐸푆푅(휆, 푡) (2.1)

Here 퐸(휆, 푡) is the irradiance at wavelength (λ) and time (t), 퐸푚푖푛 is the baseline irradiance, Δ퐸푆퐶 are the effects from the solar cycle and Δ퐸푆푅 are the effects from solar rotation. When flares are included, the equation extends to the following:

퐸(휆, 푡) = 퐸푚푖푛 + Δ퐸푆퐶(휆, 푡) + Δ퐸푆푅(휆, 푡) + Δ퐸퐺푃(휆, 푡) + Δ퐸퐼푃(휆, 푡) (2.2)

Here Δ퐸퐺푃(휆, 푡) is the gradual phase component of the flare and Δ퐸퐼푃(휆, 푡) is the impulsive component of the flare. These components are found by using the GOES XRS measurements, which have a high temporal resolution of 3 seconds. The gradual phase component uses the soft x-ray bin from GOES XRS (0.1 nm – 0.8 nm) as its proxy. It has been shown that the positive time derivative of the soft x-ray flux represents the temporal evolution of the hard x-rays quite well (Neupert 1968; Dennis and Zarro 1993). This is referred to the ‘Neupert effect’. To calculate the impulsive phase component, the positive time derivative is taken and used as its proxy.

Uncertainly estimates follow the least squares methodology and can be found in

Chamberlin et al. (2007; 2008). There are few instruments to compare FISM with and some comparisons are shown in Chamberlin et al. (2008). Results compare well with available observations, with average errors of 10% at wavelengths greater than 14 nm. There was more error at short wavelengths (< 13 nm), with average errors as large as 50%. This is due to limited observations in this region. Updated versions of FISM now include observations from the Solar

Dynamics Observatory (SDO) Extreme Ultraviolet Variability Experiment (EVE), which helps to correct this problem and improve results in the soft and hard x-ray region.

To include FISM into the WACCM and WACCM-X models, a couple of steps had to be taken. The first was to average the data to 5-minute time steps instead of 60 seconds to match the time step of the model. The second step was to bin the data into 23 wavelengths to represent the

EUV and x-ray parts of the spectrum, which includes Ly-α. The final product was irradiance binned into 23 wavelength bins with 5-minute time resolution. Ionization in the model was computed from methods described in Solomon and Qian et al. (2005).

An overview of this thesis is as follows. Chapter 2 will discuss the impacts of introducing high-energy photons and solar protons into the WACCM model. Results from simulations during

September 2005 will be presented. Chapter 3 will explain in detail the creation of the MP15 data set as well as the CMIP6 data set. WACCM simulations are compared with satellite observations are presented for validation. Lastly, chapter 4 will summarize the results of this work as well as show potential directions for future work.

Chapter 2 Effects of the September 2005 Solar Flares and Solar Proton Events on the Middle Atmosphere in WACCM

2.1 Introduction

Space weather can have considerable impacts on the earth and the near-earth environment. In addition to astronaut health and satellite degradation (Song et al., 2013), large space weather events can have impacts on the dynamics and composition of the middle and upper atmosphere. Energetic particle precipitation (EPP), the process by which charged particles from the Sun or magnetosphere impinge on the atmosphere, is one example of this (Thorne,

1980; Jackman et al., 1980; Solomon et al., 1982; Sinnhuber et al., 2012). Auroral processes cause routine precipitation of low-energy electrons and protons into the thermosphere; radiation belt processes and solar proton events (SPEs) lead to episodic precipitation of higher energy electrons and protons into the mesosphere and upper stratosphere. Precipitating particles ionize and dissociate atmospheric constituents, which results in the production of odd hydrogen (HOx =

H, OH, HO2) and reactive odd nitrogen (NOx = N, NO, NO2) (e.g., Crutzen et al., 1975; Thorne,

1980). NOx and HOx can lead to significant ozone depletion, as well as fluctuations in concentrations of other chemical species in the middle atmosphere (e.g., Crutzen et al., 1975;

Solomon et al., 1981; Jackman et al., 1995; López-Puertas et al., 2005; Randall et al., 1998;

2001; 2005).

Solar flares also influence the mesosphere and thermosphere (Donnelly et al., 1976;

Woods et al., 2004; Enell et al., 2008; Tsurutani et al., 2009). Depending on the flare's spectral

45 irradiance and class (e.g., Tsurutani et al., 2005; Le et al., 2016) and the solar zenith angle (Le et al., 2007), photoionization of the atmosphere by solar flares can occur at altitudes from approximately 80 km up through the layers of the ionosphere. The ionosphere is split into different regions. The D-region is located from approximately 60 km – 100 km, and is ionized by hard x-rays (< .1 nm) and Lyman-alpha (121.5 nm), particularly during flares (Donnelly et al.,

1976). The E-region is located between approximately 100 km and 150 km, and is ionized primarily by soft x-ray (.1 nm ≤ λ ≤ 10 nm) and EUV (> 10 nm ≤ λ ≤ 121 nm) radiation

(Donnelly et al., 1976). The top layer is the F-region, which is ionized primarily by radiation in the EUV spectral range between 10 nm and 100 nm.

In recent years, various global climate models have been used to investigate the effects of

SPEs (e.g., Jackman et al., 2005; López-Puertas et al., 2005; Funke et al., 2011). This paper examines the atmospheric effects of SPEs and solar flares using the National Center for

Atmospheric Research (NCAR) Whole Atmosphere Community Climate Model (WACCM). The focus is the first two weeks of September 2005, when 11 X-class flares and 20 M-class flares erupted from the Sun. Additionally, two medium-sized SPEs occurred at the Earth, peaking on

10 September and on 15 September (Gopalswamy et al., 2006).

WACCM is a high-top model with a vertical domain that ranges from the surface to approximately 140 km. As such, it allows the study of the effects of solar flares and solar proton events on the middle and upper atmosphere. Climate models that include the surface through the upper atmosphere are typically run with a temporal resolution that is too coarse to fully capture phenomena like solar flares. For example, WACCM generally runs with a 30-minute time step.

In addition, for long-term climate simulations, WACCM uses daily mean solar irradiance and energetic particle forcing (e.g., Matthes et al., 2017). Given that solar flares have typical

46 lifetimes of minutes to hours, the standard model temporal resolution and input cadence are clearly too coarse to accurately study their effects. To address this, a newly developed solar flare dataset was created using output from the Flare Irradiance Spectral Model (FISM), a sophisticated solar model with extremely high time and spectral resolution (Chamberlin et al.,

2007, 2008). FISM has a 3-second time cadence with 200 wavelengths between 0.5 nm and 200 nm. The FISM model was used to build a dataset that has a 5-minute temporal resolution with irradiances across 23 wavelengths in the x-ray and UV spectral ranges. In addition,

Geostationary Operating Environmental Satellite (GOES) data was used to calculate hourly ionization from the SPEs (Jackman et al., 1980 and Vitt & Jackman, 1996), rather than using daily averages as is used the standard model configuration.

The energy for solar flares and coronal mass ejections (CMEs), with which many SPEs are associated, originates from entangled magnetic fields created from both convective plasma within the Sun and circulating plasma near the photosphere (Bastian, 2010). Flares are often characterized by three stages. In the first stage (or pre-phase), continuous restructuring and destabilization of the magnetic field results in a build-up of a bipolar region in the solar corona prior to a flare; filament eruption also has been noted to occur during this stage (Shibata et al.,

2011). The second stage is referred to as the impulsive phase, and it is characterized by hard x- ray emission from proton and relativistic electron acceleration (Heyvaerts et al., 1977). The X-17 flare that was released on 7 September showed strong flux during this phase. The last stage is referred to as the main phase or gradual phase. The gradual phase can last up to an hour or longer and is dominated by release of soft x-rays and EUV radiation. Because of the duration of this phase, flares that have strong flux during this phase can be highly geo-effective. The X-6.2 flare

47 that erupted on 9 September showed strong flux in the soft x-ray and EUV spectral regions during this phase.

The location of the eruption on the Sun has a large impact on the flare's geo-effectiveness

(Donnelly et al., 1976; Tsurutani et al., 2009; Qian et al., 2010; Le et al., 2011). In particular, the further from the solar disk center the flare erupts, the more solar atmosphere the photons have to penetrate in order to reach the Earth. The solar atmosphere is a strong absorber of EUV radiation, so EUV photons emerging nearer to the limb are much more heavily attenuated than

EUV photons emerging from the center. However, the solar atmosphere is optically thin with respect to x-rays, so the x-ray flux is less affected by where the flare is located (Donnelly et al.,

1976; Qian et al., 2011); this is particularly true for X-class flares (Le et al., 2011). The two flares examined here erupted from slightly different locations on the limb of the Sun. According to the National Geophysical Data Center (retrieved from https://www.ngdc.noaa.gov/stp/solar/solarflares.html), the X-17 flare erupted at 89° central meridian distance (CMD), while the X-6.2 flare erupted at 62° CMD. As discussed below, differences in the flare spectra were partly a result of the different locations of the flare eruptions.

The most important wavelengths for influencing the upper mesosphere and lower thermosphere (MLT) are the x-rays and EUV (Fuller-Rowell & Solomon 2010; Solomon & Qian

2005). The peak ionization altitude for soft X-ray (.1 nm - 10 nm) is about 110 km, whereas the peak ionization altitude for EUV (10 nm - 120 nm) is about 160 km. Peak ionization from longer wavelength radiation (> 120 nm) ranges from the upper stratosphere near 50 km to the lower thermosphere near 150 km. The enhanced photoionization during a flare causes, among other

48 things, a dramatic increase in electron concentration, an altitude-dependent temperature increase, and an increase of odd nitrogen production in the MLT.

Several observational studies have attempted to quantify the atmospheric impact of solar flares. Donnelly et al. (1976) was one of the first to evaluate the effects of x-ray and EUV emission on the ionosphere. They found that the EUV impulsive phase of the flare created large amounts of ionization in the E and F regions. They also found that this response was dependent on the location of the flare eruption on the Sun. The main phase, which was dominated by soft x- rays and EUV radiation, caused ionization mostly in the E-region between 100 and 130 km. In

2001, a new approach for investigating flares was developed. Afraimovich et al. (2001a) used

GPS technology to quantify the total electron content (TEC) variability from solar flares. TEC is a measure of the total column number density of electrons, where one TEC unit (TECU) is 1016 electrons/m2. In a follow-up study, Afraimovich et al. (2001b) investigated two solar flares that occurred on 23 September 1998 and 29 July 1999. Both were small, M-class flares with peak observed fluxes of 7.1x10-5 Wm-2 and 5.1x10-5 Wm-2 in the 0.1-0.8 nm channel of GOES. They found that TEC changed by 0.4 TECU for the September flare and 0.5 TECU for the July flare for one line-of-sight (LOS) measurement. Since the development of this technique, other authors have used similar methods to detect the ionosphere’s response to flares. Liu et al. (2006) looked at several strong flares during the early 2000s using the GPS TEC observations. They found a linear relationship between the flare intensity at the x-ray wavelengths and the variation in TEC.

Tsurutani et al. (2005) looked at the Halloween storms, which took place in late October and early November in 2003, as well as the Bastille Day event in July of 2000. They found that the spectral variability of the flares has a large effect on the change in TEC. The 28 October 2003 flare was measured at an X-17, whereas the 4 November 2003 flare was measured at a minimum

49 of X-28. Despite this, the X-17 flare produced a TEC increase of ~25 TECU, while the X-28 (at least) flare caused an increase of only 5-7 TECU. The differences were attributed to the fact that the X-28 flare showed a larger hard x-ray increase, but a smaller EUV and soft x-ray increase than the X-17 flare due to the respective flare locations on the Sun. The X-28 flare erupted from the right-hand limb of the Sun, whereas the X-17 flare was more centralized, resulting in a high flux in the EUV and soft x-ray.

Several ionosphere/thermosphere modeling studies have been performed to investigate the effects of flares on the thermosphere and ionosphere. Meier et al. (2002) and Huba et al.

(2005) investigated ionospheric effects from the 14 July 2000 “Bastille Day” flare. Using the

SAMI2 model, Meier et al. (2002) calculated 40% increases in F-region electron density from a

50% increase in EUV radiation. Using the SAMI3 model, Huba et al. (2005) calculated increases of ~7 TECU in the mid and low latitudes, which compared reasonably well with GPS and

TOPEX data. Pawlowski and Ridley (2011) modeled two flares during the Halloween storms on

28 October and 6 November in 2003. Using the Global Ionosphere Thermosphere Model, they found large thermosphere density increases on both the dayside and night side of Earth. They also showed that most of the energy from flares is deposited below 200 km. When the energy is absorbed at these altitudes, the thermosphere expands upward, bringing the density increases upward with it. This results in large density increases below 400 km. Huang et al. (2013) used the Thermosphere-Ionosphere-Electrodynamics General Circulation Model (TIE-GCM) and included FISM. They showed that the flare spectrum has a large influence on the thermosphere and ionosphere response. The EUV response to the 28 October 2003 flare was largest in the upper thermosphere (~400 km), while the global solar heating was strongest for the x-ray wavelengths.

CMEs often accompany large solar flares. In fact, 90% of X-class flares are associated with CME eruptions (Schrijver 2009). Although both events are driven by variations of magnetic field strength, CMEs accelerate the overlying plasma in the solar corona, causing large fluxes of energetic particles to be launched into space. CMEs often begin when magnetic flux tubes form in the solar corona; the CME is released when these tubes erupt. The plasma contained within the tubes can be accelerated to speeds as high as 2000 km/sec in directions perpendicular to the magnetic field, launching energetic particles into the heliosphere (Hudson et al., 2010).

Sometimes these particles are directed toward the Earth and cause SPEs, as occurred in

September of 2005. The influence of SPEs on the atmosphere has been investigated thoroughly over the last few decades from both a modeling and observational perspective. As reviewed by

Jackman et al. (2006), SPEs have been observed to cause large increases in odd hydrogen and odd nitrogen in the polar middle atmosphere, which can lead to significant ozone loss. Jackman et al. (2008) included the September 2005 time period in a study that investigated the short and medium-term effects of SPEs. Using WACCM3 (WACCM version 3), they found that 1.5 gigamoles of total odd nitrogen (NOy) was created from 7-17 September 2005, which makes this

th the 17 strongest SPE period, in terms of NOy production, since 1963 (Jackman et al., 2008;

2014).

2.2 Model Description

WACCM is a high-top configuration of the National Center for Atmospheric Research

(NCAR) Community Earth System Model (CESM) (Hurrell et al., 2013). The ‘high-top’ refers to the uppermost level of the atmospheric component, which extends into the thermosphere, making WACCM ideal for studying atmospheric effects from EPP. WACCM version 4 was used

51 in this study and followed the protocols defined for the Chemistry-Climate Model Initiative

(CCMI, Eyring et al., 2013). WACCM is based on the Community Atmospheric Model, version

4 (CAM4), which is on a hybrid-sigma coordinate system with 66 pressure levels from the surface extending to 5.1 x 10-6 hPa. The vertical resolution is 1.1 km in the troposphere, 1.1-1.4 km in the stratosphere below 30 km, 1.75 km from 30-50 km, and 3.5 km above 65 km. The horizontal resolution of WACCM4 is 1.9° x 2.5°. The chemistry module in WACCM derives from the Model for Ozone and Related Chemical Tracers, version 3 (MOZART3), which is discussed in detail by Kinnison et al. (2007), with updates described in Solomon et al. (2015).

The simulations used in this study applied the ‘Specified Dynamics’ WACCM (SD-WACCM) mode (Lamarque et al., 2012; Brakebusch et al., 2013). SD-WACCM uses reanalysis data from the Modern-Era Retrospective Analysis for Research and Applications to nudge the meteorological conditions in the model to match actual meteorological conditions for the dates of the simulation (Rienecker et al., 2011). That is, at every time step SD-WACCM calculates new wind and temperature fields by taking 99% of the calculated model data and 1% of the meteorology data. The nudging in this study occurs from the surface to approximately 40 km.

The nudging is linearly reduced from 40-50 km, where the model becomes free-running (not nudged). Using SD-WACCM also increases the number of pressure levels from 66 to 88.

Typically, WACCM is run with a 30-minute time step, but for the simulations presented here, the model time step was reduced to 5-minutes to better resolve the flare effects. Sensitivity studies showed that in the absence of flares or SPEs, reducing the time step to 5 minutes had negligible impact on the middle atmosphere. Switching to the 5-minute time step led to some increases in clouds in the troposphere, mainly in the equatorial region, but differences between the two model simulations above the tropopause were insignificant.

2.3 Numerical Experiments

The two primary flares that were investigated in this study differed substantially in both their measured strength and their influence on the atmosphere. The flare that occurred on 7

September was measured at X-17, whereas the flare on 9 September was only X-6.2. The first flare was located on the far western limb of the Sun at 89° CMD, while the X-6.2 flare erupted at

62° CMD. In order to best capture the effects from the solar flares, both WACCM and the input data needed to have high temporal resolution.

For this study, four simulations were performed (Table 1). The first was a baseline simulation where no forcing from SPEs or flares was applied. The second simulation was forced just by SPEs and the third was forced just by flares. The fourth was the fully forced simulation that contained both SPEs and flares. All simulations were also forced with daily aurora input, based on the parameterized auroral oval of Roble and Ridley (1987). Input values of proton fluxes for the SPEs were taken from GOES measurements. A detailed description of how the proton fluxes are input into the model is available in Vitt and Jackman, (1996). In previous work, hourly proton fluxes were binned into daily averages before the ion production rates were calculated (Jackman et al., 2008; Jackman et al., 2009). In the simulations with solar proton

53 forcing presented here, the calculation of ion production rates used hourly, not daily, proton fluxes. Figure 1 shows the difference between the hourly and daily ion production rates. The reduced time cadence (Figure 1a) shows a more detailed and accurate portrayal of the SPEs than the daily time cadence.

The reduced time step in the model was coupled with high time resolution solar data from

FISM (Chamberlin et al., 2008). FISM is a solar spectral irradiance empirical model that uses observational data from the GOES X-ray measurements, the Thermosphere Ionosphere

Mesosphere Energetics and Dynamics (TIMED) Solar Extreme ultraviolet Experiment (SEE)

(Woods et al., 1994) and the Solar Radiation and Climate Experiment (SORCE) Solar Stellar

Irradiance Comparison Experiment (SOLSTICE) (Rottman et al., 1993). It includes wavelengths from 0.5 to 200 nm at 1-nm spectral resolution with a 3-second time cadence. A more complete discussion of the FISM algorithm for both daily cadence and flare effects can be found in

Chamberlin et al. (2008). For the flare simulations included in this study, the FISM data was processed into 5-minute averages at 23 wavelengths. For the other simulations, the short-lived flares were removed and the FISM data was reduced to daily averages. The importance of the high time cadence of the flare input is illustrated in Figure 2. This figure shows the irradiance in

54 the wavelength bin centered on 1.3 nm during the flare period using the 5-minute (black solid) and daily averaged (purple dashed) time cadences, as well as the daily averaged irradiance with the flare removed (red dashed). The 1.3 nm, soft x-ray wavelength was chosen in Figure 2 because flares emit copious amounts of radiation at this wavelength during both the impulsive phase and the main phase. Thus, it is a good indicator for flare strength and flare duration. It can be seen clearly that the 5-minute cadence is needed to fully capture the irradiance variability during the flares.

Figure 3 shows the irradiance at various wavelengths for both flares as calculated by

FISM. It is clear from Figures 3a and 3b that the 7 September flare (black line) was a faster- rising, shorter duration flare compared to the 9 September flare (red line). The 7 September flare began at 17:17 UT, with a flare peak at 17:40 UT. Despite being larger according to the flare class (X-17), the EUV flux enhancement (panels c and d) from the 7 September flare was smaller than that of the 9 September flare, which began at 19:13 UT, since the X-17 flare was located

55 closer to the solar limb than the X-6.2 flare. At 85.5 nm (not shown), the irradiance enhancement from the 9 September flare over the first 4 hours was 25% more than that of the 7 September flare. Much of the EUV energy is deposited at altitudes above the top boundary of WACCM

(Solomon et al., 2005). However, for wavelengths longer than 100 nm, which are absorbed within the WACCM vertical domain, the flux from the second flare is much larger than from the first flare. This should result in larger impacts on electron density and odd nitrogen production in the 9 September flare simulation.

2.4 Results and Discussion

An increase in electron concentration is one of the primary consequences when solar flare radiation reaches the Earth. Significant photoionization occurs as photons strike the dayside of the Earth, with the ionization altitude dependent on the flare spectrum. Figure 4 illustrates the difference in electron number density between the simulation that included flares and the

56 baseline simulation from 1-15 September 2005. The top panels (a-c) show the values for the baseline simulation, with the middle and bottom panels showing the electron number density differences (flare minus baseline) in units of percentage and # electrons/m3, respectively. Panels

(a), (d) and (g) show the northern hemisphere (NH) polar average (60°N – 90°N), panels (b), (e) and (h) show the southern hemisphere (SH) polar average (60°S – 90°S), and panels (c), (f) and

(i) show the equatorial average (30°S – 30°N). Individual flares are evident in the plots as vertical stripes between 7 September and late on 9 September. Electron increases in the equatorial region (Panels f and i) are larger than in either of the polar regions, which is expected since the solar zenith angle is smaller in the equatorial region. Similarly, electron increases in the late NH polar summer (left column) are higher than in the late SH polar winter (middle column).

Since the month of interest is September, the NH receives more light from the Sun than the SH, resulting in higher increases in the NH. The maximum number density differences, all of which are associated with the large X-17 flare on 7 September, are 7.4 x 1010 e-/m3 in the polar NH at approximately 110 km, 2.5 x 1010 e-/m3 in the polar SH also at 110 km, and 1.1 x 1011 e-/m3 in the equatorial region at 100 km. The electron number density increases are quite brief in the lower thermosphere (above ~100 km), and last only a few hours; but they reach differences of well over 500% relative to the baseline simulation. The electron increases in the mesosphere

(~60-90 km) persist longer than in the thermosphere. It is likely that this is caused by faster recombination of the dissociated electrons in the thermosphere than in the mesosphere. One of the primary recombination pathways for electrons is through combination with NO+. NO+ concentrations increase significantly in the thermosphere, which results in more rapid recombination above the mesopause; in the mesosphere, there is less NO+ available for recombination, resulting in longer lifetimes of the free electrons. Small increases on the order of

108 e-/m3 can be seen for several days after the solar flares between 70 and 100 km. Electron number densities remained elevated for less time after the strongest flare, which occurred on 7

September, than after the second strongest flare, which occurred on 9 September. The 9

September flare had a longer decay time during the gradual phase, resulting in significantly more

EUV radiation; consequently, electron density increases lasted longer. Figure 4 also shows small differences ranging from 0-1% above 100 km after 10 September. Since no flares are represented in the model at this time, we attribute these subtle differences to internal variability in the model.

Figure 5 shows Mercator maps of the effects of the 7 September X-class flare for electron number density (top), temperature (rows 2 and 3), and NOx mixing ratio (bottom). All plots

58 correspond to a pressure level of 10-4 hPa (~110 km), except for those in row 3, which correspond to 2.7x10-5 hPa (~125 km). We found 10-4 hPa to be ideal for illustrating the variations during solar flares because this is a level where effects from all variables are seen, but we added 2.7 x 10-5 hPa for temperature because the effects are more prominent. The left column shows the baseline simulation values for each variable at 17 UT on 7 September. Other columns show 30-minute averaged differences between the flare and baseline simulations in two-hour time increments beginning at t0=18 UT.

As expected, once the flare peaks, large electron number density differences can be seen exclusively on the dayside of the Earth (inside the terminator). Increases relative to the baseline of between 100 and 500% are found during the peak on the dayside. After two hours, the electron density difference is reduced by a factor of 10, to 10-50% increase from the background.

After four hours, the electron density difference is reduced to less than a 10% increase relative to the background at high solar zenith angles, with slightly higher increases approaching the terminator. Note that electron differences only appear inside the terminator, which is evidence of the short lifetimes of electrons at this altitude. Unlike the electron enhancements, the average temperature differences from the baseline simulation are highest beginning an hour after the peak of the flare and remain elevated for a few hours at the longitudes where the flare originally hit the Earth. These temperature differences, which are caused by the heating of the thermosphere due to enhanced ionization, are altitude dependent. At 110 km, the temperature differences are 2-

3 K. At 125 km, they are as high as 10K, while at both altitudes the effects remain at the location of the initial flare. In the NH near 70°-80°N latitude and 280°-340°E longitude, there are small temperature decreases of between 2 and 8 K (depending on altitude) that we believe are due to model internal variability. The bottom panels show the NOx differences from the baseline

59 simulation for the same time period. The peak NOx production follows the similar pattern as the temperature changes, with a slight delay from the peak of the flare of an hour. Like temperature,

NOx produced by the flare is not transported significantly beyond the production region during the first three hours. The weakened ionization following the peak of the flare correspondingly produces considerably less NOx.

Figure 6 is analogous to Figure 5, but for the X-6.2 flare that occurred on 9 September.

The baseline simulation background values on the left panels are shown for 19 UT; the

60 difference plots begin at t0=20:30 UT. Many of the differences are similar in character to, but larger than, those in Figure 5, except initial electron production. The top plots show that the electron density enhancements are slightly smaller than for the previous flare during the peak, but last significantly longer than in the previous flare. Four hours past the peak of the flare, we still see enhancements of 100% relative to the baseline simulation. As explained below, that the second flare was longer in duration led to more overall ionization by the X-6.2 flare on 9

September than by the X-17 flare on 7 September. The peak temperature differences at 125 km, which reach as high as 12 K, occur two hours after the flare’s peak. This increase is 20% higher than in the previous flare. However, at the lower altitude of 110 km, the temperature difference is not as recognizable. The temperature increases are more localized and not as widespread as in the previous flare. The NOx differences show larger increases that last longer than in the first flare; NOx differences greater than 100% from the baseline simulation last 4 hours after the flare begins.

Figure 7 shows the time series of NOx (a, c) and HOx (b, d) mixing ratio differences (%) between the forced simulations and the baseline simulation for the NH polar region from 60°N-

90°N. Figures 7a and 7b give the differences between the SPE-forced simulation and the baseline simulation, and Figures 7c and 7d show the differences between the flare-forced simulation and the baseline simulation. The larger of the SPEs began on 8 September and peaked on 10

September. The consequent NOx and HOx production was most significant in a relative sense from about 1 hPa to 0.01 hPa. NOx differences greater than 1000% persisted in the lower mesosphere near 0.1 hPa for several hours. The NOx-rich air descended over the next several weeks, resulting in elevated values until about 20 October at altitudes as low as 3 hPa. HOx

62 increases also maximized near 0.1 hPa, reaching values of about 30%; as expected, the HOx enhancements were much shorter-lived than the NOx increases. Flare-induced NOx increases of about 60% to 80% are evident in the NH polar region around 0.01 hPa, but these differences are much smaller than the increases caused by the SPEs; there were no significant flare-induced HOx increases. That the SPE effects were much larger than the flare effects at 60°N-90°N is expected, since particle precipitation effects maximize in the polar region and flare effects maximize at lower latitudes.

SPE- and flare-induced changes in NOx and HOx in the SH polar region (60°S-90°S;

Figure 8), are similar in character to those in the NH. However, the SPE-induced NOx increases are smaller in the SH than in the NH, and the SPE-induced HOx increases are larger in the SH than in the NH. In addition, the flare-induced NOx increases appear to be delayed relative to the flare-induced ionization. The different % increases in the NH vs. the SH polar regions can be attributed to differences in the baseline NOx and HOx values at the end of the winter in the SH and summer in the NH (see, e.g., Jackman et al., 2014). The apparent delay in the NOx increase is the result of the time required for air with flare-induced NOx enhancements to descend into regions where these enhancements are significant with respect to the baseline NOx mixing ratios.

As expected, no SPE influences are evident in the equatorial region from 30°S to 30°N

(Figure 9), because the precipitating protons are trapped by the magnetic field lines of Earth, and thus are restricted to geomagnetic latitudes above 60° in WACCM. The solar flare influence on

NOx, however, is stronger here than in the polar region, with NOx increases of up to 150% relative to the baseline simulation. Since flare influences only occur on the sunlit side of the

Earth, it is expected that the tropics would show the largest NOx production. The solar flare influence on HOx in the equatorial region was insignificant relative to the baseline HOx mixing ratios. This is expected because of the very high baseline HOx (primarily H) mixing ratios above

0.01 hPa (>5000 ppmv; not shown), and the short lifetime of HOx below 0.01 hPa in sunlight

(e.g., Pickett et al., 2006). Figure 9 appears to suggest that the 9 September flare had a larger impact on NOx production than the 7 September flare, but as explained next, this is partially due to the additive effects of relatively long-lived NOx from previous flares.

To explain why the initial NOx enhancements after the 9 September flare were larger than after the 7 September flare, even though the opposite was true of the electron density enhancements, Figure 10 compares time series of electron density and NOx enhancements.

Figures 10a and 10b show the electron number density differences between the flare simulations and the baseline simulations. Figure 10a includes the geographic region from 7°S - 7°N latitude and 250°E - 275°E longitude, where the electron density enhancements after the 7 September flare maximized, while Figure 10b includes the geographic region from 7°S - 7°N latitude and

210°E - 235°E longitude, where the electron density enhancements after the 9 September flare maximized. These locations are indicated by black boxes on Figure 5 and Figure 6 for the 7

September and 9 September flares, respectively. Figures 10c and 10d show the same regions but for NOx mixing ratio differences instead. All panels pertain to an altitude of approximately 110 km. The electron differences are very brief and show no cumulative effects; that is, after both flares the electron density increases returned to background values before subsequent flares arrived. However, the NOx mixing ratio differences show that there are in fact cumulative effects whereby the NOx mixing ratios after the 9 September flare include contributions from previous flares. That is, the NOx mixing ratios had not decayed back to their baseline values before the 9

September flare hit. When calculated relative to the values immediately preceding each flare, the

4 4 maximum change in NOx was 1.7x10 ppbv for the 7 September flare, and 1.4x10 ppbv for the 9

September flare. Thus, like the electron densities, the initial NOx production was smaller after the 9 September flare.

Quantifying the overall flare impacts on the atmosphere requires examination not only of the initial effects, but also the integrated effects. The average electron density enhancements inside the geographic regions indicated by the black boxes in Figures 5 and 6, calculated in 30- minute increments over the first 6 hours after the beginning of each of the flares, were 3.79 x

1010 e-/m3 for the 7 September flare and 4.82 x 1010 e-/m3 for the 9 September flare. This indicates that because the 9 September flare was longer in duration, the total ionization was in fact larger than after the 7 September flare. This result emphasizes the importance of the different spectral characteristics of the two flares. Since the X-6.2 flare on 9 September was stronger during the gradual phase, the enhanced photoionization lasted longer, and produced more electrons overall than the X-17 flare on 7 September.

Vertical profiles of the individual NOx and HOx components are shown in Figure 11. The top panels (a-c) show differences (fully-forced minus baseline) for the various NOx constituents, and the bottom panels (d-f) show differences for the HOx constituents. Polar averages from

60°N-90°N (a, d) and 60°S-90°S (b, e) correspond to 10 September at 12 UT, near the peak of the SPE. Equatorial averages from 30°S to 30°N (c, f) correspond to 9 September at 21 UT, near the peak of the solar flare on that date. Increases in atomic nitrogen and NO2 are apparent in all regions, but the vast majority of the increase in NOx takes the form of NO; in fact, the NO and

NOx curves are nearly indistinguishable in the NH polar region and in the equatorial region. NOx partitioning shifts toward NO2 below ~1 hPa in the SH polar region, which is just emerging from the polar night. Indeed, the shift in partitioning from NO to NO2 increases monotonically with latitude from 60°S to 90°S (not shown), since the pole is still in total darkness at this time of year. SPE-induced HOx increases take the form of OH throughout much of the ionization region near 0.1 hPa, although HO2 becomes more significant near 1 hPa in the SH polar region. HOx partitioning shifts to atomic hydrogen at the highest altitudes at all latitudes, consistent with

Pickett et al. (2006).

Figure 12 explores the effects of the flares and SPEs on ozone at latitudes from 60°N-

90°N (a) and 60°S-90°S (b); no significant changes in ozone were discernable in the equatorial region where flare effects are simulated in other species. Ozone decreases of 40-50% in the NH polar region and over 80% in the SH polar region were caused by the SPEs. The ozone decreases were short-lived and correspond primarily to increases in HOx, particularly OH (indicated by the white contour), consistent with well-known HOx chemistry (Jackman et al., 2008). The model simulations did not reveal any temperature changes that could be ascribed to the HOx-induced ozone decreases (not shown). Although SPE-induced NOx enhancements persisted into October in both the NH and SH, the descending NOx did not reach stratospheric altitudes where NOx catalytic loss dominates ozone loss. Thus the EPP indirect effect (Randall et al., 2006) was inconsequential for stratospheric ozone loss, a result that is not surprising since the mesospheric polar vortex in September-October is just forming in the NH and breaking up in the SH.

To evaluate WACCM calculations of NO production by the SPEs and flares, model NO was compared to observations from the Atmospheric Chemistry Experiment Fourier Transform

Spectrometer (ACE-FTS) on the Canadian SCISAT-1 satellite (Bernath et al., 2005). ACE-FTS is a solar occultation instrument that measures NO with a vertical resolution of 3-4 km over an altitude range of 6-107 km (Kerzenmacher et al., 2008; Sheese et al., 2016). Figure 13 compares daily mean vertical profiles of WACCM NO mixing ratios (Figure 13a), calculated for the ACE measurement times and locations in the NH, to ACE NO mixing ratio profiles (Figure 13b); note that all of the ACE profiles at these NH latitudes were satellite sunrise occultations. Generally, the NO mixing ratios from WACCM are similar to the ACE mixing ratios. However, WACCM underestimates the mixing ratios relative to ACE during these solar events. On 10 September during the peak of the SPE, ACE shows a larger NO increase near 60-90 km than what WACCM calculates, with differences of approximately 100 ppbv. At higher altitudes (90 – 100 km) differences as high as 5000 ppbv are evident. This result could be due to the omission of medium energy electrons in the model (e.g., Peck et al., 2015), and is consistent with results from previous studies (Randall et al., 2015).

2.5 Conclusion The investigation presented here utilizes new, high-time-cadence solar spectral input in the WACCM coupled chemistry climate model to investigate the effects on the middle atmosphere of solar flares and SPEs during September 2005. Using a 5-min time cadence irradiance input and hourly SPE input, individual WACCM simulations were performed to isolate both solar proton and solar flare effects on middle atmosphere electron production, temperature, and mixing ratios of NOx, HOx, and ozone. The flares caused electron number

71 densities to increase by more than 500% in the thermosphere, with largest increases in the equatorial region as expected. Temperature changes were primarily concentrated in the equatorial regions near the sub-solar point, with a maximum of 10 K at 125 km during the X-17 flare on 7 September, and 12 K at 125 km during the X-6.2 flare on 9 September.

The flares that were investigated in this study differed substantially in both their measured strength and their influence on the atmosphere. The X-17 flare was a fast-rise flare with a shorter decay time than the X-6.2 flare, which had a prolonged decay phase. In addition, the X-17 flare on 7 September was located on the far western limb of the Sun, while the X-6.2 flare on 9 September erupted slightly closer to the center. Consequently, the two flares had remarkably different spectral characteristics. For instance, the smaller flare caused a larger increase of irradiance in the EUV but smaller increase in the x-ray. Wavelengths between 20 nm and 100 nm tend to deposit energy above WACCM’s vertical domain. The first flare had a stronger peak flux in wavelengths less than 20 nm, but the time-integrated flux over the following four hours was close to the same. The peak fluxes at wavelengths longer than 20 nm were higher in the 9 September flare. The 7 September flare caused a larger initial increase in electron density as well as NOx production at the peak of the flare. In contrast, the time- integrated effects of the 9 September flare were larger. Spectral variability has a large impact on the flare influence on the atmosphere. Effects by X-ray flux are well represented in WACCM, but some of the EUV effects will be at altitudes above 140 km and thus cannot be simulated by

WACCM. This suggests that simulations of the smaller flare on 9 September would show even more atmospheric impacts if the EUV radiation outside the model’s domain could be represented. The upcoming release of WACCM-X will solve this problem by expanding the model vertical domain to 500 km.

The simulations show that the SPEs had a much larger impact on NOx and HOx in the polar mesosphere than any of the flares. The SPEs caused increases of more than 1000% in NOx in the polar mesosphere, with enhancements of at least 25% persisting into mid-October. SPE- induced HOx increases reached 90% in the SH polar mesosphere and about 30% in the NH polar mesosphere, but were short-lived; the hemispheric differences were attributed primarily to differences in the background HOx values. Short-lived ozone loss of up to ~50% in the NH polar lower mesosphere and over 80% in the SH polar lower mesosphere was attributed to the SPE- produced HOx. No changes in temperature could be ascribed to this ozone loss. Longer-lived ozone loss from the EPP indirect effect was not evident in the simulations because the SPE- produced NOx did not descend to the stratospheric altitudes most relevant for the NOx catalytic cycle. Had the SPEs studied in this work occurred in mid-winter instead of in September, significant ozone loss from the EPP indirect effect would have been more likely (e.g., Jackman et al., 2008). Since solar protons precipitate in the polar regions, the SPEs did not perturb the equatorial mesosphere. Nor did they significantly impact the thermosphere, even in the polar regions, since their energy was deposited primarily in the mesosphere.

Relative to the SPEs, the flares caused more modest and shorter-lived NOx increases in the polar mesosphere; flare-induced NOx enhancements of 60-80% decayed to less than 25% by around 20 September. The flares caused NOx mixing ratio increases of several thousand ppbv in both the tropical and polar thermosphere, but these increases represented relative changes of less than 25%. The work presented here demonstrates for the first time the ability of a fully coupled climate model such as WACCM to simulate solar flare effects on the middle atmosphere using very high time resolution for the flare spectral input. While there were significant flare-induced

73 increases in electron density and odd nitrogen in the mesosphere, the flares did not significantly impact mesospheric HOx or ozone, and did not affect the stratosphere.

Chapter 3 Atmospheric Effects of >30 keV Energetic Electron Precipitation during the SH winter of 2003

3.1. Introduction

Space weather’s influence on the earth’s atmosphere and near-earth environment has attracted significant attention since the satellite era. Solar protons and high-energy electrons can be very dangerous to astronaut health and can damage expensive space instrumentation (Parsons and Townsend 2000; Moreno-Villanueva et al., 2017). In addition, charged particles can enter the atmosphere and disturb the chemistry of the middle and upper atmosphere through energetic particle precipitation (EPP). Atmospheric effects of solar protons and auroral electrons (energy <

30 keV) have been studied extensively over the last few decades (Jackman et al., 1995; 2004;

2008; 2009; Orsolini et al., 2005; López-Puertas et al., 2005b; Funke et al., 2011; Roble and

Rees, 1977; Crutzen, 1979). Particularly in the past several years, more attention has been paid to medium energy electron (MEE; ~30-1000 keV) and high-energy electron (HEE; > 1 MeV) precipitation influences on the atmosphere (Andersson et al., 2018; Seppälä et al., 2018; Clilverd et al., 2013; Newnham et al., 2018; Verronen et al., 2015). Evaluating two different energetic electron precipitation (EEP) data sets for inclusion in global climate models, both of which include MEE precipitation inferred from the Medium Energy Proton and Electron Detector

(MEPED) instruments, is the focus of this paper. One of the EEP data sets incorporates data from

75 only the 0° MEPED telescope, while the other incorporates data from both the 0° and 90° telescopes.

EEP causes ionization of atmospheric constituents, with electrons of higher energies reaching deeper into the atmosphere. The primary region of energy deposition is ~60-90 km in altitude for MEE precipitation, and lower for HEE precipitation (e.g., Fang et al., 2008, 2010).

The ionization leads to significant increases in reactive odd nitrogen (NOx = N + NO + NO2) and odd hydrogen (HOx = H + OH + HO2) (Thorne, 1980; Jackman et al., 1980; Solomon et al.,

1982; Solomon et al., 1981). Both of these chemical families are involved in catalytic ozone destruction, making them important for the chemistry and the dynamics of the middle atmosphere. In addition to direct production of NOx and HOx by EEP, the indirect effect of EEP

(Randall et al., 2006; 2007) also dictates the extent to which EEP affects the middle atmosphere.

During the polar winter, when limited light increases the lifetime of the NOx via photolysis, it descends from the thermosphere and mesosphere into the stratosphere. Furthermore, mesospheric air that was depleted in ozone via EEP-HOx (Andersson et al., 2014) may descend into the stratosphere in the polar winter. In order to simulate accurately the atmospheric effects of EEP, it is important for models to correctly simulate both EEP-induced production of HOx and NOx as well as descent rates in the polar winter.

One critical challenge in quantifying EEP effects is determining the spectral energy flux of electrons that precipitate into the atmosphere (Clilverd et al., 2010; Rodger et al., 2013;

Turner et al., 2012). Currently, the longest continuous data set of precipitating MEE comes from the observations made by the MEPED instruments onboard the Polar Orbiting Environmental

Satellite (POES) constellation, which measure MEE fluxes in the bounce loss cone (e.g., Rodger et al., 2010). Codrescu et al. (1997) studied the influence of MEE precipitation on the

76 mesosphere and thermosphere using the Thermosphere Ionosphere Mesosphere Electrodynamics

General Circulation Model and MEPED data. The simulations showed that MEE precipitation could influence NOx and HOx concentrations, leading to ozone depletion in the mesosphere.

However, ionization rates used in that study were uncertain since the MEPED electron and proton detectors suffer from multiple issues. Examples are: cross contamination between the electron and proton telescopes (Asikainen and Mursula, 2013; Lam et al., 2010; Rodger et al.,

2010; Yando et al., 2011; Peck et al., 2015), radiation damage to the proton sensors (Asikainen et al., 2012), that the telescopes do not sample the entire loss cone (Rodger et al., 2013), and that the instruments provide only highly integrated measures of the spectral flux (Evans and Greer,

2004). Because of these limitations, and the limited time frame over which MEPED data are available (van de Kamp et al., 2016), global climate models have historically excluded MEE and

HEE from their specification of EEP, and therefore underestimate the amount of EEP-induced ionization in the atmosphere (e.g., Randall et al., 2015). Without accounting for all ionization sources, simulated chemistry changes in the middle atmosphere would be incomplete and underestimated.

Several independent efforts have been made to correct the errors associated with the

MEPED instruments, and as a result several corrected electron flux data sets have been created

(Lam et al., 2010; Asikainen and Mursula, 2013; Green et al., 2013; Peck et al., 2015; Nesse-

Tyssøy et al., 2016; Asikainen and Ruopsa, 2019). The study presented here compares results from three different simulations of the National Center for Atmospheric Research (NCAR)

Whole Atmosphere Community Climate Model (WACCM), a fully coupled chemistry climate model. All three simulations included auroral electron precipitation, but differed in their inclusion of MEE precipitation. A baseline simulation omitted MEEs. A second simulation

77 specified ion production rates (IPRs) according to the Coupled Model Intercomparison Project version 6 (CMIP6) recommended forcing data set described in van de Kamp et al. (2016) and

Matthes et al. (2017); this data set includes MEE IPRs derived from measurements made by the

0° MEPED telescope. In a third simulation, IPRs for MEE were specified by a modified version of the data set described by Peck et al. (2015), which used measurements from both the 0° and

90° MEPED telescopes. Results from all three simulations are compared with each other and with satellite observations for the southern hemisphere (SH) winter of 2003.

WACCM is an appropriate model for studying EEP impacts on the atmosphere because the model vertical domain extends to about 140 km. It has been used in previous studies of the atmospheric effects of EEP. Randall et al. (2015) showed that WACCM underestimated the amount of NOx descending to the stratosphere in the Arctic springtime of 2004, attributing the underestimate to a combination of insufficient EEP (only auroral electrons were included in the model used) and inadequate descent rates. Eyring et al. (2016) reported the development of a long-term EEP data set for CMIP6 model comparisons, which was derived from the MEPED measurements using the Ap magnetic index as a proxy (van de Kamp et al., 2016; Matthes et al.,

2017). Andersson et al. (2018) incorporated the CMIP6 EEP data set into a free-running version of WACCM to investigate polar ozone losses from MEE precipitation. They found MEE precipitation to have a significant impact on ozone loss in both the mesosphere and stratosphere.

Another recent study used simulations from WACCM-D (WACCM with D-region chemistry) both with and without medium energy electrons as specified by Nesse Tyssøy et al. (2016) to simulate the April 2010 EEP event (Smith-Johnsen et al., 2018). They found that using D-region chemistry with WACCM improved the direct production of NO versus WACCM without D- region chemistry. Strong correlations between the model and observations were found in the

78 lower thermosphere and in the middle mesosphere; however, near 90-110 km the model underestimated the NO.

The SH winter of 2003 was chosen for the investigation described here for two reasons.

First, since the SH shows less variability in polar winter dynamics than the northern hemisphere, simulations of SH winter dynamical conditions are likely to be more robust, facilitating investigations of chemical effects. Second, in May through August of 2003 elevated levels of geomagnetic activity resulted in significant EEP, which makes this an ideal winter for evaluating simulations of EEP effects. The following section discusses the details of the WACCM model and the simulations performed as well as how the EEP data sets are generated. Section 3 describes the results of the simulations and comparisons with observations. Finally, section 4 summarizes and discusses the conclusions of the study.

3.2. Numerical Simulations and Observations

WACCM is a high-top configuration of the NCAR Community Earth System Model

(CESM) (Hurrell et al., 2013). The ‘high-top’ refers to the uppermost level of the atmospheric component, which extends into the thermosphere. WACCM version 4 (Marsh et al., 2013) was used in this study, following the protocols defined for the Chemistry-Climate Model Initiative

(CCMI, Eyring et al., 2013). WACCM4 is based on the Community Atmosphere Model, version

4 (CAM4), and uses a hybrid-sigma coordinate system with 66 pressure levels from the surface extending to 5.1 x 10-6 hPa. The horizontal resolution of WACCM4 is 1.9° lat x 2.5° lon. The chemistry module in WACCM derives from the Model for Ozone and Related Chemical Tracers, version 3 (MOZART3), which is discussed in detail by Kinnison et al. (2007), with updates described in Solomon et al. (2015). The simulations used in this study applied the ‘Specified

Dynamics’ WACCM (SD-WACCM) mode (Lamarque et al., 2012; Brakebusch et al., 2013).

SD-WACCM uses reanalysis data from the Modern-Era Retrospective Analysis for Research and

Applications (MERRA) (Rienecker et al., 2011) to nudge the meteorological conditions in the model to match actual meteorological conditions for the dates of the simulation. That is, at every time step (30-minutes) SD-WACCM calculates new wind and temperature fields by taking 99% of the calculated model data and 1% of the meteorology data. The nudging in this study occurs from the surface to approximately 40 km. The nudging is linearly reduced from 40-50 km, where the model becomes free running (not nudged). SD-WACCM also increases the number of levels from 66 to 88.

To help improve the chemistry in the mesosphere, the D-region chemistry module is adopted in this work, which adds 307 reactions in addition to the default MOZART chemistry

(Verronen et al., 2016). Many of these reactions are from water cluster ions that are common in the mesosphere. The mesosphere has small quantities of water up through the mesopause. Water cluster chemistry can have a significant impact on odd nitrogen in the mesosphere, but is not included in the default model (Verronen et al., 2016). As shown by Newnham et al. (2018) and

Smith-Johnsen et al. (2018), SD-WACCM-D describes the mesosphere and lower thermosphere chemistry during geomagnetically active times more accurately than regular WACCM. For simplicity, for the remainder of the paper we refer to the model used, as WACCM, but it should be understood that this is the specified dynamics version of WACCM4 with D-region chemistry.

The MEPED instruments (Evans and Greer, 2004) used in this study are part of the

Space Environment Monitor-2 (SEM-2) platforms that fly aboard POES and the European Space

Agency MetOp satellites. The satellites have near circular orbits at about 850 km. Each MEPED instrument consists of two proton telescopes and two electron telescopes. Each telescope pair has

80 a telescope that points 9° away from the zenith (“0°” detector), while the other points 9° away from the anti-ram direction (“90°” detector). Each telescope has a 30° field of view with a combined 2-second sampling time using both detectors. The proton telescopes have 6 broadband energy channels, labeled P1 to P6 (Evans and Greer, 2004). The highest energy channel

(P6), which will be discussed more below, measures protons with energies greater than 6.9 MeV.

The proton telescopes have cobalt magnets designed to bring incident electrons into an aluminum bin, preventing them from reaching the detector. The electron telescopes have three integrated energy channels, which measure the number of electrons with energies greater than 30 keV (labeled E1), 100 keV (E2), and 300 keV (E3). The electron telescopes have a nickel-foil cover to prevent low energy protons from entering the detector. The electrons that reach the detector collide into a silicon sheet that is designed to absorb electrons with energies up to 2.5

MeV. Using MEPED data to specify EEP ionization rates requires consideration of several previously documented problems. Cross contamination between the proton and electron detectors causes false detections that must be removed from the electron and proton data (Lam et al., 2010; Green et al., 2013; Rodger et al., 2010; Peck et al., 2015), and radiation damage has been shown to affect older satellites (Asikainen and Mursula, 2012). Radiation damage is one source of error that is present in both the CMIP6 EPP ionization rates as well as the energetic electron flux data set created according to the method of Peck et al. (2015). This source of error can become significant in just a few years after the instrument is launched (Asikainen and

Mursula 2013; Asikainen and Ruopsa 2019; Nesse- Tyssøy et al., 2016). In addition, the instruments do not sample the entire bounce loss cone (Rodger et al., 2013), so accurately calculating the total precipitating flux requires estimating the MEE pitch angle dependence

(Nesse-Tyssøy et al., 2016).

The two data sets used in this work, from CMIP6 (van de Kamp et al., 2016) and from

Peck et al. (2015), were chosen because they represent improvements on previous MEPED- based electron precipitation data sets, are publicly available, and are easily incorporated into

WACCM. The CMIP6 method of calculating EEP is based on measurements from the 0°

MEPED detectors, because the 0° detectors measure electrons within the bounce loss cone, while the 90° detectors measure both trapped and bounce loss cone electrons at middle and high latitudes (Nesse-Tyssøy et al., 2016). The method described in Lam et al. (2010) is applied to remove proton contamination from the electron measurements. The corrected electron counts are converted to directional fluxes using the conversion method described in Evans and Greer

(2004). In cases where the electron directional flux in the E1 channel is less than 250 electrons cm-2 sr-1 s-1, all channels are set to zero. The corrected directional fluxes are given with 0.5 L- shell resolution and one-day temporal resolution across all magnetic local times. To calculate the spectral flux from the three electron channels (E1-E3), a power law spectrum is assumed. As explained in van de Kamp et al. (2016), the power law assumption breaks down at energies greater than 1 MeV, so HEE precipitation is not included in the CMIP6 calculations.

After calculating the spectral flux for the time period from 2002-2012 as just described, the data are used to derive a relationship between the fluxes and the Ap index for that time period. This relationship is then used to infer the precipitating spectral flux for any given value of the Ap index, and these Ap-based fluxes are then used to compute vertical profiles of ionization rates via the mono-energetic method of Fang et al. (2010). The ionization rate calculations use temperatures and densities from an empirical model of the atmosphere (Picone et al., 2012), rather than the WACCM atmosphere. The resulting data set contains daily ionization rates binned by L-shell. An additional step was taken in this study to interpolate the

CMIP6 ionization rates to the WACCM grid by converting L-shell to magnetic latitude and finally to geographic latitude.

The second EEP data set considered by this study is a modified version of the EEP data set of Peck et al. (2015); it is referred to hereinafter as the MP15 data set. Like the CMIP6 calculation, the MP15 data set is also based on POES MEPED data; but there are several notable differences between the CMIP6 and MP15 calculations. The first difference is the incorporation of MEPED measurements from both the 0° and 90° electron detector telescopes, which serves to broaden the observed pitch angle distribution (Nesse-Tyssøy et al., 2016). Based on previous work (e.g., Gu et al., 2011; Vampola, 1998), for MP15 it is assumed that the pitch angle (α) dependence of the precipitating particle flux varies as sin(α), with no particle flux at 0° pitch angle and maximum particle flux at 90° pitch angle. Second, to remove proton contamination, inversion methods from O’Brien and Morley (2011) were used instead of the method described in Lam et al. (2010). Details of the inversion method can be found in Peck et al. (2015). Third, since the P6 channel on the proton telescope is intended to record protons with energies >6.9

MeV but also responds to relativistic electrons, in the absence of protons in the P5 channel

(which records protons in the range 2.5-6.9 MeV), the P6 channel is a good proxy for highly energetic electrons (Rodger et al. 2010). Therefore, Peck et al. (2015) use the P6 channel to create an additional virtual electron channel (E4) that quantifies electrons with energies (>1

MeV). Thus, all four channels (E1-E3 plus E4) are used in the calculation of MP15 differential flux values, and the energetic electron flux spectra extend to 10 MeV instead of 1 MeV. Fourth, rather than assuming that the differential electron flux follows a power law, Peck et al. (2015) fit a combined spectrum of a relativistic Maxwellian, double Maxwellian, power law, and energy exponential to the MEPED electron channel measurements.

Three improvements to the method of Peck et al. (2015) are included for our current investigation. Peck et al. (2015) included electron counts in all L-shells from the MEPED detectors. However, at very low geomagnetic latitudes MEPED is sensitive to both drift cone and trapped electrons as well as to bounce loss cone electrons. The majority of the electrons that exist at L-shells less than 2 are most likely trapped electrons that will not precipitate into the atmosphere (Rodger et al., 2013). Therefore, to improve the MEPED data set for this investigation, any electron counts below L-shell 2 were removed prior to processing the data. In addition, to be consistent with the CMIP6 data set, electron counts from L-shells larger than 10 were also removed. Lastly, after converting count rates to directional fluxes, those measurements with fewer than 250 electrons cm-2 sr-1 s-1 were removed from the data set, since this is close to the noise floor of the MEPED instruments; this is in agreement with the CMIP6 method (van de

Kamp et al., 2016).

In addition to the differences in electron flux calculations, the CMIP6 and MP15 simulations also differ in their approach for undertaking ionization rate calculations. Instead of calculating ionization rates offline using a separate atmospheric model as in the CMIP6 simulation, for the MP15 simulation the electron flux spectra are interpolated from the original

MEPED measurement locations to the WACCM grid points. In order to include enough observations to create a robust map, each daily map on the WACCM grid includes five days of

MP15 electron flux spectra, centered on the day of interest. The results are hemispheric maps of electron flux spectra that are used as input to WACCM, and ionization rates are computed self- consistently within WACCM using the model temperatures and densities, following Fang et al.

(2010). Since ionization rates based on Fang et al. (2010) have not yet been validated for electrons greater than 1 MeV, fluxes in energy bins higher than 1 MeV were removed before

84 calculating ionization rates. Thus, neither the CMIP6 nor the MP15 EEP data sets include HEE precipitation. The MP15 maps allow for 3D ionization rate profiles, as compared to the CMIP6 data, which are zonally averaged on each L-shell.

WACCM results are compared with several different satellite instrument observations.

The Michelson Interferometer for Passive Atmospheric Sounding (MIPAS; Fischer et al., 2008) was operational from 2002-2012 and measured several minor atmospheric constituents as well as temperature using atmospheric limb emissions from 4.15 μm to 14.6 μm. As an infrared emission instrument, MIPAS sampled globally each day. This study uses NOx (NO+NO2) from the

MIPAS version V5H data set (Funke et al., 2014). For the time period of interest here, NOx is available at altitudes from the clouds’ top to 72 km, with a vertical resolution of 3 km in the stratosphere and 5-10 km in the mesosphere. The Halogen Occultation Experiment (HALOE;

Russell et al., 1993) is one of four solar occultation instruments used in this study. It measured temperature and several trace gases including NO and NO2 from 1991-2005. Like other solar occultation instruments in low earth orbit, it provided ~15 vertical profiles around two different latitude circles on any given day, at spacecraft sunrise and sunset. HALOE Version 19 NOx data is used here (Gordley et al., 1996; see also Randall et al., 2002). The second solar occultation instrument employed is the Polar Ozone and Aerosol Measurement (POAM) III. POAM III measured vertical profiles of NO2 between 20 and 60 km, from 1998-2005. The POAM NO2 data used here is version 6.0 (Randall et al., 2002), which has a vertical resolution of 1-2 km between

22 km and 37 km in altitude, increasing to 3 km near 40 km and 7 km at 45 km. The third solar occultation instrument is the Stratospheric Aerosol and Gas Experiment (SAGE) III on the

Meteor-3M spacecraft, which operated from 2002-2005. SAGE III measured ozone, aerosols and

NO2 density profiles from cloud top to 40 km with a vertical resolution of 1 km (Rault et al.,

2004). The SAGE III NO2 data product version 3.0 is utilized here.

The three simulations conducted for this investigation are referred to as the baseline,

CMIP6, and MP15 simulations. The baseline simulation included auroral electrons, but no MEE or HEE. The CMIP6 simulation included the CMIP6 MEE ionization rates, and the MP15 simulation used spectral fluxes as described above, with ionization rates calculated inline in

WACCM. All model simulations were run from January through September 2003 and were forced with daily aurora input based on the parameterized auroral oval of Roble and Ridley

(1987). All three simulations included ionization from solar protons. Input values of proton fluxes for the solar proton events (SPEs) were taken from GOES measurements. A detailed description of how the proton fluxes are input into the model is available in Vitt and Jackman

(1996). In previous work, hourly proton fluxes were binned into daily averages before the ion production rates were calculated (Jackman et al., 2008; 2009). In the simulations presented here, the ion production rates from SPEs were hourly, not daily. During the SH winter of 2003, there were very minor SPEs on 29 May, 31 May, and 19 June, but no major events.

3.3. Results and Discussion

Figure 1 gives a snapshot of the ion production rates from the CMIP6 and MP15 data sets on two days in May of 2003 at altitudes of 90 km and 75 km. This was during a period of enhanced geomagnetic activity. Both simulations show a similar geographic distribution at these altitudes, with a displacement toward Australia due to the location of the magnetic pole.

However, peak ionization rates in the MP15 case are more than a factor of three higher than in the CMIP6 calculation. At lower altitudes, the MP15 shows higher ionization rates as well and

86 the peak ionization rates are geographically different between the two data sets. Particularly noteworthy is the extension of high ionization rates to more equatorward latitudes in MP15 than in CMIP6 at lower altitudes; this has an important effect on the model results, and will be discussed more below.

Figure 2 shows a time series of the ionization rate vertical profiles from the CMIP6 and

MP15 data sets during the SH winter from March through September 2003. The rates are

87 averaged from 50°S to 80°S geographic latitude in Figure 2a and 2b, and averaged from 30°S to

50°S in Figure 2d and 2e. The differences between the two are shown in Figure 2c and 2f

(MP15-CMIP6). At high latitudes, the timing of ionization events is similar in both data sets, although there are some differences that we attribute to the fact that the CMIP6 data set is parameterized according to the Ap index, so it does not necessarily reproduce the MEPED variations precisely. As noted by van de Kamp et al. (2016), errors associated with the CMIP6

EEP ionization rates can reach a factor of 10. It is clear from the figure that overall the MP15 ionization rates are larger than the CMIP6 rates, and that the MP15 ionization extends to lower altitudes, as quantified in panels c and f. The CMIP6 simulation resulted in very little ionization at 30°S-50°S (panel d), whereas the MP15 simulation shows significant ionization at these latitudes during geomagnetically active time periods. The larger rates are primarily due to the

MP15 use of both the 0° and 90° MEPED telescopes, which extends the latitude extent farther equatorward than using just the 0° telescope. Another interesting feature is the difference in altitudes where the peak ionization occurs. The peak CMIP6 ionization rates occur at nearly the same altitude during all periods, whereas the MP15 peak ionization altitude varies with the event.

For example, the altitude of the MP15 peak ionization on 1 June is near a pressure level of 0.01 hPa (~80 km), while the 1 August peak is at 0.001 hPa (~96 km). This is likely due to differences in the spectral fits to the MEPED electron channel data, which would change the fluxes of electrons at higher energies (300 keV – 1 MeV). As mentioned in the previous section, MP15 includes a combination of analytical functions to compute the spectral flux, whereas the CMIP6

88 spectral flux calculation uses a power law.

To show the impact of MEE precipitation on OH, Figure 3 shows OH mixing ratio differences between the baseline simulation and the model simulations that were forced with the

CMIP6 (Figure 3a) and MP15 (Figure 3b) data sets. The plots show the differences as forced minus baseline; for brevity, we refer to the simulations that include MEE ionization as “forced”, but it should be understood that even the baseline simulation includes auroral electrons and

SPEs. Results were averaged over latitudes from 50°S-80°S and are shown for the months of

March through September. Over-plotted on each panel are the MEE ionization rates at 0.01 hPa from the respective simulations. We chose to show this altitude because this is where the largest change in OH is seen. Tick marks on the horizontal axes denote the first day of each month. The

CMIP6 simulation shows relatively little OH increase from MEE precipitation, whereas the

MP15 simulation shows large increases during periods of electron precipitation. The black lines indicate that the timing of the OH increases coincided with MEE ionization. OH increases of at least 2 ppbv can be seen during periods of elevated EEP in the MP15 simulation. The odd hydrogen lifetime in the mesosphere is on the order of hours (Solomon et al., 1983), so as expected, Figure 3 shows no evidence of the OH being transported to lower altitudes. The MP15 simulation did show some moderate ozone loss in the SH polar mesosphere from the OH production during this time period (not shown); ozone mixing ratio losses of 0.5 - 0.75 ppmv were found at 0.05 hPa, and coincided with the altitudes and times of OH increases.

Figure 4 shows NOx mixing ratios from the MIPAS instrument (a), the baseline simulation (b), the CMIP6 simulation (c), and the MP15 simulation (d), during the SH 2003 winter. All panels show averages between 70°S and 90°S, and the model was sampled at the satellite observation times and locations. The model results were interpolated to an altitude grid with a 1 km vertical resolution using model geopotential height. The black contours are included

90 for guidance, and denote the 16 ppbv and 64 ppbv levels from the MIPAS data. Areas of white in

Figure 4a indicate either missing MIPAS data (so these areas are also shown as white in the numerical simulations), or MIPAS data with errors greater than 200%. All simulations show a

“tongue” of descending NOx throughout the winter, as expected from the observations. This is unambiguously identified as EPP-NOx, since EPP is the only source of NOx in the mesosphere and lower thermosphere (MLT) during the polar winter. It is clear, however, that both the baseline and the CMIP6 simulations underestimate the descending NOx mixing ratios relative to

MIPAS, with the low bias apparent all the way up to 70 km. In combination with the fact that the

CMIP6 simulation agrees slightly better with MIPAS than the baseline, this suggests that the underestimate is caused at least partially by too little production of EPP-NOx, although too little descent in the MLT could also contribute. More EPP-NOx is produced in the mesosphere in the

MP15 simulation because of the higher ionization rates, so the MP15 simulation more accurately matches the observations. Even in the MP15 simulation, however, WACCM often shows a NOx deficit in the tongue, indicating that too little EPP-NOx descended into the stratosphere. That this most likely results from insufficient descent rates is supported below by an analysis of WACCM carbon monoxide (CO), which serves as a tracer of vertical transport.

To quantify the differences between the forced simulations and the observations, the percent differences in NOx mixing ratios between the WACCM simulations and the MIPAS measurements are presented in Figure 5. Areas of blue show where WACCM underestimates

NOx relative to MIPAS, whereas red regions indicate overestimations. The CMIP6 simulation

(Figure 5b) is a significant improvement over the baseline simulation (Figure 5a), but results show a systematic NOx deficit of 20% – 80% relative to MIPAS at most altitudes in May through

August. During this same time period, differences between the MP15 simulation and MIPAS are more often within ±40%, consistent with the qualitative conclusion from Figure 4 that the MP15 simulation more accurately reproduces the tongue of descending EPP-NOx. However, Figure 5 also shows large regions where all three simulations overestimate NOx compared to MIPAS.

This is particularly evident below 60 km in March and April, and above 40 km in late August and September. The former is attributed to a high NOx bias in the initial model conditions that descends with time as air is transported downward by the residual circulation. The latter appears

92 to reflect a high bias in the amount of descending EPP-NOx above the top altitude of the MIPAS measurements, which then descends with time from 60-70 km in late August into the stratosphere in September. This bias could be caused by too much production of NOx via EEP

(EEP-NOx) and/or too-rapid descent in the mesosphere and lower thermosphere (MLT), but no measurements are available to definitively distinguish between these possibilities. The high bias in late August and September is largest in MP15, followed by CMIP6 and then the baseline simulation, which suggests that the different MEE flux estimates contribute to the bias. That even the baseline simulation shows a high bias would argue that too much auroral EEP and/or too-rapid descent in the MLT also contributes to the bias. An overestimate of auroral EEP is consistent with the WACCM results for the northern hemisphere winter of 2004 reported by

Randall et al. (2015), but inconsistent with the WACCM simulations of a geomagnetic storm in

April 2010 reported by Smith-Johnsen et al. (2018). The possibility that the bias is caused by too rapid (or prolonged) descent is revisited below in the discussion of Figure 8.

As discussed above, incorporating data from both the 0° and 90° MEPED telescopes in the MP15 data set results in more EEP-NOx production at mid-latitudes than in the CMIP6 simulation. To investigate the mid-latitude effects further, Figure 6 compares MIPAS NOx

93 observations with the model results for latitudes between 40°S and 50°S. The dashed black line denotes the 16 ppbv NOx contour from the MIPAS observations, and is included for guidance.

The MIPAS data show clear evidence of EEP-NOx descending from the mesosphere into the stratosphere at mid-latitudes during the SH winter. The CMIP6 simulation shows very little descending EEP-NOx, because electron ionization using only the 0° detector would be confined primarily to more polar latitudes. The MP15 simulation more closely matches the observations, but does underestimate NOx mixing ratios in the stratosphere and mesosphere. This could indicate too little ionization at these latitudes and altitudes, and/or that there are errors in the meridional and/or vertical transport. Nevertheless, Figure 6 confirms that including the 90° telescope, and possibly better constraining the spectral distribution for energies of 300 keV –

1MeV, improves simulation results significantly for the mid-latitudes.

To broaden the comparisons to other available measurements, Figure 7 compares observations of NOx at 45 km altitude from HALOE (top row), and of NO2 at 40 km altitude from SAGE III (middle) and POAM III (bottom, all in black), with WACCM results at the corresponding locations and times from the baseline (left), CMIP6 (middle) and MP15 (right) simulations (all in red). The latitudes of the measurements are plotted in blue and referenced to the right vertical axis. In March to September of 2003, HALOE measurement locations swept rapidly through sunlit latitudes between 69°S and 77°N, resulting in observations that were only occasionally at SH latitudes influenced significantly by MEE precipitation. Figure 7 shows all

HALOE SH measurements poleward of 40°S, since these are the latitudes most relevant to the current investigation. Note that most of these measurements were between 40°S and 50°S, with

95 excursions to polar latitudes in March and early April, and an excursion to 56°S in August. As might be expected for an altitude of 45 km at mid-latitudes, there is little difference between the baseline and CMIP6 simulations, since neither includes ionization from electrons at these latitudes. All three simulations overestimate NOx mixing ratios at 40 km in March through May, reflecting a high bias that was present at the beginning of the simulations. However, by late June, when the HALOE data exhibit a substantial enhancement due to descending EPP-NOx, all three simulations underestimate the observed NOx. This is consistent with the mid-latitude MIPAS results in Figure 6, and with the early deficit in simulated descending NOx at polar latitudes shown in Figures 4 and 5. Nevertheless, the MP15 simulation, which includes electron flux at lower latitudes, more closely matches the HALOE NOx mixing ratios from mid-July through

August, once again highlighting the need to include count rates from the 90° telescope in electron flux calculations.

Moving to slightly more polar latitudes, the middle row of Figure 7 shows comparisons between the WACCM simulations and SAGE III measurements at 40 km. Since the SAGE III measurement latitudes changed slowly with time, Figure 7 shows seven-day running averages of the measurements. Only NO2, not NOx, is shown, because SAGE III did not measure NO. One can infer from the MIPAS data in Figures 4 and 6 that the increase in NO2 mixing ratios observed by SAGE III in late June and July was caused by descent of EEP-NOx to 40 km. The subsequent decline in August most likely indicates that the tongue of NOx-rich air had descended below 40 km, so NO2 at 40 km was returning to background levels at this time. Both the CMIP6 and baseline simulations show a deficit of NO2 from July through September, with maximum differences of about 2 ppbv, or 50% of the observed NO2. The MP15 simulation, however,

97 compares very well with SAGE III. The improved agreement is again attributed to the fact that

MP15 includes electron precipitation outside the polar region, which is influential at the latitudes sampled by SAGE III. Thus, in agreement with the mid-latitude MIPAS and HALOE comparisons, the SAGE III comparison confirms that including measurements from both

MEPED telescopes in electron flux calculations is important for accurate simulation of EEP effects at mid latitudes.

As shown by the bottom row in Figure 7, POAM III sampled latitudes poleward of 60°S throughout the March to September time period. Like SAGE, POAM measurement latitudes changed slowly with time, so Figure 7 shows seven-day running averages of the measurements; also like SAGE, POAM measured NO2, but not NO. The MIPAS data in Figure 4 indicate that

40 km was near the lower edge of the tongue of descending polar NOx in June and July, well within the tongue throughout most of August, and above the tongue in September. Thus, as in the

SAGE observations, the increase in POAM NO2 mixing ratios in June and July is indicative of the descent of EEP-NOx from higher altitudes down to 40 km. The decline in August-September is attributed mostly to the fact that, as noted above, the tongue of EEP-NOx descended below 40 km in September, so polar NOx mixing ratios at 40 km were declining to their background values at this time. There might also be a contribution from the change in latitude of the POAM measurement sampling, from ~71°S in early August to 88°S on 22 September. MIPAS data confirm that NO2 mixing ratios at 40 km decreased slightly toward the pole in September (not shown), so it is expected that the POAM sampling excursion would result in a decline in NO2.

Consistent with the MIPAS NOx comparisons from 70°S-90°S in Figures 4 and 5, all of the WACCM simulations underestimate the increase at the POAM measurement locations in

June and July, but the MP15 simulation comes closest to reproducing it. Maximum NO2 mixing

98 ratios of 3.5 ppbv were measured by POAM in early August, and only the MP15 simulation shows values up to 3 ppbv; maximum mixing ratios in the baseline and CMIP6 simulations are only ~2 and 2.5 ppbv, respectively. It thus appears that the amount of EEP-NOx that descends to

40 km in the MP15 simulation is sufficient, but that the timing is delayed, possibly because of too little descent to 40 km in June and early July. As noted above, POAM NO2 mixing ratios declined in August-September to less than 2 ppbv, whereas the NO2 mixing ratios in all three simulations remain relatively constant (baseline and CMIP6) or decline just slightly (MP15) after reaching their maximum values. That all of the simulations show too little decrease in 40-km

NOx mixing ratios in September suggests that the cause is most likely unrelated to the EEP specification, and is more likely due to errors in MERRA-forced WACCM transport caused either by too much replenishment, or too little removal, of NOx-rich air at 40 km at the POAM measurement locations. As noted above, the MIPAS comparisons in Figures 4 and 5 suggest that there is too much descent of NOx-rich air from the mesosphere into the stratosphere in late

August and September in the WACCM simulations.

To evaluate vertical transport in WACCM, simulated CO was compared with MIPAS

CO, since CO is a tracer that can be used as a proxy for descent in the mesosphere and upper stratosphere during the winter (Allen et al., 1999; Harvey et al., 2015). Figure 8 shows CO mixing ratios from MIPAS (a) and WACCM (b), and the percent differences (c), averaged over

70°S-90°S in March through September of 2003. Results from only the baseline simulation are shown; but as expected, CO mixing ratios are similar in all three simulations. The black contour lines are guides that denote the 32 ppbv, 256 ppbv, and 1024 ppbv contour lines from MIPAS.

Qualitatively, the WACCM and MIPAS CO mixing ratios are morphologically similar. Except for a week in early May, however, the MIPAS CO contours in March through early June are

99 steeper than the WACCM CO contours, which indicates that WACCM descent is too slow at this time. While this should lead to negative differences between WACCM and MIPAS, the systematic high bias in WACCM at the beginning of the time series leads to the positive differences in Figure 8c throughout March at most altitudes, and also in April through June at sequentially lower altitudes as the CO at 70 km and above descends with time. In April and May below 60 km, the vertical gradient in CO steepens in WACCM (i.e., the color contours in Figure

8b become compressed), because simulated descent rates at the lower altitudes decrease relative to those at the higher altitudes. This does not appear to occur in the observations (Figure 8a). In addition, the MIPAS data indicate strong downwelling below 50 km in mid-May that is not captured in the simulation. Together, these differences lead to the swath of negative differences in Figure 8c that begins in late March and descends with time through June. Descent rates above

40 km from mid-June through September cannot be inferred from Figure 8; but a continued underestimate of descent rates early in the winter would be consistent with the early winter NOx deficits discussed above. WACCM CO mixing ratios in August and September are higher than

MIPAS mixing ratios throughout the polar stratosphere, which appears from Figure 8b to result from too much CO descent. If this is indeed indicative of descent that is too prolonged in

WACCM, it could explain the high bias in NOx mixing ratios in the MP15 simulation in

September, and the lack of NO2 decline at 40 km in the comparisons in Figure 7.

100

3.4. Conclusions

This work compares SD-WACCM simulations of the 2003 SH winter using two different estimates of MEE ionization rates. In the “CMIP6” SD-WACCM simulation, the publicly available CMIP6 EEP ionization rates (Eyring et al., 2016) were prescribed. The CMIP6 ionization rates are derived from an Ap-index-based parameterization of measurements from the

MEPED 0° electron detector telescope (van de Kamp et al., 2016). In the “MP15” SD-WACCM simulation, MEE spectral fluxes from an improved version of the MEE data set described by

Peck et al. (2015) were input to SD-WACCM, and ionization rates were calculated within SD-

WACCM. The MP15 electron fluxes are based on measurements from both the MEPED 0° and

90° electron detector telescopes. Both the CMIP6 and MP15 SD-WACCM simulations included precipitating electron energies up to 1 MeV. However, the analytical fits used to convert spectrally integrated MEPED electron fluxes to differential spectral fluxes used different functional forms, and the MP15 fits extended to energies as high as 10 MeV (even though the spectra were truncated at 1 MeV for inclusion in SD-WACCM). A baseline simulation was also performed, which included auroral electrons but no MEE precipitation. Ionization rate

101 comparisons show that the MP15 electron fluxes correspond to significantly more ionization than the CMIP6 fluxes during the 2003 SH winter. This is probably due to the utilization of both

MEPED telescopes in the MP15 calculation, with some contribution from the different spectral fit used in the MP15 flux calculation, which would increase the flux of electrons on the high energy tail of the distribution.

Results from the three WACCM simulations were compared with observations from the

MIPAS infrared emission spectrometer as well as from the HALOE, SAGE III, and POAM III solar occultation instruments. NOx mixing ratios from the CMIP6 and MP15 simulations agree better with observations than NOx mixing ratios from the baseline simulation. Compared to the

CMIP6 simulation, the MP15 simulation shows significantly more EEP-NOx descending into the stratosphere, which is in better overall agreement with observations. Particularly noteworthy is the substantially improved agreement in the MP15 simulation at mid-latitudes (40°S-50°S).

Although the MP15 simulation reproduces the main body of descending EEP-NOx reasonably well, some significant disagreements with the observations are also apparent. For instance, NOx mixing ratios at the lower altitude edge of the descending body of EEP-NOx are underestimated in all three simulations relative to the MIPAS data. Comparisons between

WACCM and MIPAS CO, a tracer of vertical motion, suggest that in the specified dynamics version of WACCM descent rates in the lower mesosphere and upper stratosphere are too low early in the winter, resulting in too little descending EEP-NOx. On the other hand, the MIPAS comparisons suggest that there is too much EEP-NOx above 70 km in late August in the MP15 simulation, leading to a high bias as the excess NOx descends into the stratosphere; this is also true for the CMIP6 simulation, but to a much smaller degree. This might indicate too much production of EEP-NOx in the mesosphere and lower thermosphere; but it is also consistent with

102 descent rates that are too strong in late winter, as suggested by the comparisons of modeled and measured values of CO. At mid-latitudes, all three models underestimated EPP-NOx descending into the stratosphere. Additional ionization at energies > 1 MeV could be contributing to the lack of EEP-NOx shown in the model simulations here. Future work will examine the contribution of

HEE precipitation during geomagnetic activity.

To summarize, the WACCM simulation that includes MP15 electron fluxes reproduces the descent of EEP-NOx from the mesosphere into the stratosphere during the 2003 SH winter more accurately than the simulation that includes the CMIP6 electron fluxes. The CMIP6 simulation performs better than the baseline simulation, but it underestimates ionization rates, resulting in less NOx and OH production from EEP. Although not investigated in this study, underestimating NOx and HOx would also lead to underestimating ozone destruction. An updated version of the CMIP6 electron data set has been produced recently (van de Kamp et al., 2018), and should be included for future comparisons with other electron data sets. However, a key advantage of the MP15 data set over CMIP6 is that MP15 incorporates measurements from both the 0° and 90° MEPED telescopes. Thus, the MP15 calculations include a wider range of precipitating electron pitch angles and energies. Another advantage is that the MP15 multi- function spectral fit used to convert the MEPED integrated channel data to differential fluxes better constrains the high-energy tail of the EEP distribution than a simple power law, as used in the CMIP6 calculations. The model-measurement comparisons in this study confirm that including both electron detector telescopes, which means considering the full range of pitch angles, is critical for accurate simulations of the effects of EEP on the atmosphere, particularly at mid-latitudes. This study confirms that a comprehensive understanding of the impacts of EEP on the atmosphere requires consideration of the full range of pitch angles of precipitating electrons,

103 as well as the full range of energies, as first suggested more than three decades ago by Baker et al. (1987).

104

Chapter 4

Conclusions and Future Work

The emphasis on understanding variability of the middle and upper atmosphere due to space weather events over the past decades has answered many questions. Nonetheless, many important questions remain unanswered. Insufficient coverage from observational data, coarse model resolution and lack of understanding of certain physical processes allow some questions to remain a mystery. The work presented here focuses on energetic electron and high-energy photon influences on the atmosphere using WACCM, a fully coupled global climate model.

Progress over the past two decades has achieved success with integrating proton ionization into models, while electrons and high-energy photons have proved more difficult.

The first study here attempted to see how well WACCM simulates solar flares. We found from Pettit et al. (2018) that impacts from solar flares do have a moderate effect on MLT chemistry in sunlit regions, with large impacts to electron densities in the D and E regions of the ionosphere. We also found that the spectral variance between flares has a pronounced influence on the geo-effectiveness, regardless of the flare strength. The spectral variance depends primarily on the type of flare erupting (impulsive or gradual) as well as the location on the Sun of the eruption. For the flares studied here, a strong impulsive flare was compared with a weaker gradual flare. Interestingly, we found the effects of the weaker flare to be more pronounced in the thermosphere than the stronger flare. Flares that produce higher fluxes in the EUV have larger impacts above the D-region, while flares with large x-ray signatures have greater impacts on the D-region. In the case study here, the strong, impulsive flare showed large x-ray signatures and created more significant changes in the mesosphere than the weaker flare. The weaker flare,

105 nonetheless, showed larger changes to the thermosphere from the prolonged and strong EUV emission. Unfortunately, WACCM is unable to reach altitudes high enough to capture the

September 2005 flares’ full effects. In order to capture the full effects solar flares have on the atmosphere, a model with a higher ceiling would have to be utilized.

The third chapter of this thesis demonstrated the importance of including HEE into global climate models. Recent studies have shown the need for including energetic electrons into models to capture correctly chemistry effects from geomagnetic storms. Pettit et al. (2019a) focused on the 2003 SH polar winter using WACCM-D. The first simulation (baseline) only included auroral electrons, the second simulation (CMIP6) included MEE precipitation that was calculated using an Ap-model (CMIP6), and the last simulation (MP15) included both MEE and

HEE precipitation (MP15). The MP15 simulation agreed better with observations, showing that including HEE is necessary for accurate modeling of EEP effects. The simulation that included just MEE (CMIP6) performed better than the baseline simulation but still underestimated NOx mixing rations in the polar region. In the sub-auroral latitudes, the MP15 simulation performed excellent relative to the other two simulations. The CMIP6 and baseline simulations showed similar results due to neither having ionization at these latitudes.

State of the art models such as WACCM include solar protons and auroral electrons by default in simulations. The work presented here demonstrates the need to include MEE and HEE during periods of enhanced geomagnetic activity. The improvements made to the methods described in Peck et al. (2015) to create the MP15 data set provide an excellent long-term database of MEE and HEE that can be included in future model simulations. The importance of including these electrons was clearly seen during the 2003 SH polar winter. This particular winter did have significant electron precipitation, but it is not anomalous in comparison with

106 other years during solar maximum. Even during solar minimum years, the Sun’s 27-day rotation causes periodic geomagnetic activity. Therefore, the need for future model studies investigating the middle atmosphere to include MEE and HEE is important for simulating the correct chemistry. The MP15 data set can provide the MEE and HEE for future simulations.

We also found that during intense solar flares, high-energy photons cause significant impacts to the MLT region. While the impact at polar regions during the winter is minimal, the sunlit regions see substantial ionization. FISM provides a way to include these photons into models to capture the ionization that occurs during solar flares. Prior to the work presented here, no simulations used a fully coupled, surface to model top GCM to simulate effects from solar flares. Thus, utilizing both the FISM data set as well as the MP15 data set, long-term transient runs can be performed that include solar protons, auroral electrons, both MEE and HEE, as well as high-energy photons to get a complete picture of space weather impacts on the middle and upper atmosphere.

Several challenges await future work studying space weather impacts on the atmosphere.

One of the main difficulties with modeling the mesosphere and thermosphere in models is the inability to constrain dynamics. Using SDWACCM, dynamics can be constrained in the model up through the stratosphere. However, above the stratopause the model becomes free-running, which can result in large differences between the real-world dynamics and the model dynamics.

One method that has confronted this issue is the use of WACCM using the Data Assimilation

Research Testbed (DART). WACCM-DART uses several additional forms of observations including the Aura Microwave Limb Sounder and the Sounding of the Atmosphere using

Broadband Emission Radiometry to constrain dynamics above the stratosphere (Pedatella et al.,

2014). This is likely the best method for limiting dynamical variability in regions where little

107 observational data, and no reanalysis products, are available. Future model simulations that investigate the polar region during dynamically active time periods, such as during and after

SSWs, should employ the use of WACCM-DART.

Another difficulty confronting scientists is the availability of long-term observations with good spatial resolution. Observations of middle atmosphere chemistry and temperatures are needed to properly validate model simulations. Additionally, calculating electron fluxes that precipitate into the atmosphere has been an ongoing struggle due to limited electron observations. The MP15 data set introduced in this thesis uses interpolation methods in order to create 3D maps for WACCM. This methodology induces large errors due to limited satellite observations. FISM also suffers from limited observations since its database relies on only 39 large flares. The second version of FISM looks to expand this database significantly. However, the problem of incomplete data in the x-ray and EUV spectral regions still remains.

Despite these challenges, future work can still answer many unknown questions. Three recommendations are made here:

1) Long-term simulations should be performed to test the climatological impacts MEE and HEE could have on the polar and sub polar regions. The SEM-2 instrument aboard the POES has data back to 1999. This means the MP15 data set begins in 1999. Running transient simulations from 1999 through the present using the MP15 data set would help understand the long-term impacts of MEE and HEE. 2) Constrain the dynamics using WACCM-DART during years of enhanced dynamic activity in the polar region. Capturing the correct descent rates from the mesopause through the stratosphere is fundamental for accurately simulating the impacts from energetic particle ionization. Using WACCM-DART for the 2004 NH winter would allow for a great testbed validating how well the MP15 data set is performing.

108

3) Use the updated version of FISM during various periods of intense solar flares (Halloween storms, September 2005, September 2017, etc.) and compare with GPS TEC data to validate the new version of FISM.

109

Bibliography

Allen, D. R., Stanford, J. L., López-Valverde, M. A., Nakamura, N., Lary, D. J., Douglass, A. R., et al. (1999). Observations of middle atmosphere CO from the UARS ISAMS during the early northern winter 1991/92. Journal of Atmospheric Sciences, 56, 563-583. https://doi:10.1175/1520-0469(1999)056<0563

Alves, M. V., Echer, E. and Gonzalez, W. D. (2006). Geoeffectiveness of corotating interaction regions as measured by Dst index. Journal of Geophysical Research, 111, A07S05. https://doi:10.1029/2005JA011379

Andersson, M. E.,Verronen, P. T., Rodger, C. J., Clilverd, M. A., Seppälä, A. (2014). Missing driver in the Sun–Earth connection from energetic electron precipitation impacts mesospheric ozone. Nature Communications, 5, 5197. https://doi.org/10.1038/ncomms6197

Andersson, M. E., P. T. Verronen, D. R. Marsh, S.-M. Päivärinta, and J. M. C. Plane (2016). WACCM-D { Improved modeling of nitric acid and active chlorine during energetic particle precipitation. Journal of Geophysical Research: Atmospheres, 121, 10,328{10,341, https://doi.org/10.1002/2015JD024173

Andersson, M. E., Verronen, P. T.,Marsh, D. R., Seppälä, A.,Päivärinta, S.-M., Rodger, C. J.,… van de K amp, M. (2018). Polar ozone response to energetic particleprecipitation over decadal timescales: The role of medium-energy electrons. Journal of Geophysical Research: Atmospheres, 123, 607–622. https://doi.org/10.1002/2017JD027605

Andrews, D. G., J. R. Holton, and C. B. Leovy (1987), Middle Atmosphere Dynamics, Academic Press, San Diego, California.

Afraimovich, E. L., Altynsev, A. T., Grechnev, V. V., & Leonovich, L. A. (2001). Ionospheric effects of the solar flares deduced from global GPS network data. Advances in Space Research, 27, 1333–1338. https://doi.org/10.1016/S0273-1177(01)00172-7

Afraimovich, E.L., Altyntsev, A.T., Kosogorov, E.A., Larina, N.S., & Leonovich L. A. (2001). Ionospheric effects of the solar flares of September 23, 1998 and July 29, 1999 as deduced from global GPS network data. Journal of Atmospheric and Solar-Terrestrial Physics, 63, 1841-1849. https://doi.org/10.1016/S1364-6826(01)00060-8

Asikainen, T., Mursula, K., and Maliniemi., V (2012). Correction of detector noise and recalibration of NOAA/MEPED energetic proton fluxes. Journal of Geophysical Research, 117, A09204. https://doi:10.1029/2012JA017593

Asikainen, T., and K. Mursula (2013). Correcting the NOAA/MEPED energetic electron fluxes for detector efficiency and proton contamination. Journal of Geophysical Research: Space Physics, 118, 6500–6510. https://doi:10.1002/jgra.50584

110

Asikainen, T., and M. Ruopsa. (2019). New Homogeneous Composite Of Energetic Electron Fluxes From POES: 2. Intercalibration of SEM‐1 and SEM‐2, Journal of Geophysical Research: Space Physics, 124, 1203-1221. https://doi.org/10.1029/2018JA026214

Babcock (1961). The topology of the Sun’s Magnetic Field and the 22-year cycle. Astrophysical Journal, 133, 572-587.

Baker, D. N., J. B. Blake, D. J. Gorney, P. R. Higbie, R. W. Klebesadel, and J. H. King (1987). Highly relativistic magnetospheric electrons: A role in coupling to the middle atmosphere. Geophysical Research Letters, 14, 1027-1030, https://doi:10.1029/GL014i010p01027

Bastian T. (2010). Radiative signatures of energetic particles. In C. J. Schrijver and G. L. Siscoe (Eds.), Heliophysics: Space Storms and Radiation: Causes and Effects (pp. 79-123). Cambridge: Cambridge University Press.

Bernath, P. F., McElroy T., Abrams M. C., Boone C. D., Butler M., Camy-eyret C., … Zou J. (2005). Atmospheric Chemistry Experiment (ACE): Mission overview. Geophysical Research Letters, 32, L15S01. https://doi.org/10.1029/2005GL022386

Brakebusch, M., Randall C. E., Kinnison D. E., Tilmes S., Santee M. L., & Manney G. L. (2013). Evaluation of Whole Atmosphere Community Climate Model simulations of ozone during Arctic winter 2004–2005. Journal of Geophysical Research: Atmospheres, 118, 2673– 2688. https://doi.org/10.1002/jgrd.50226

Burch, J. L., Precipitation of low-energy electrons at high latitudes: Effects of interplanetary magnetic field and dipole tilt angle. Journal of Geophysical Research, 77, 6696, 1972

Burgess D. (1995). Collisionless Shocks. In M.G. Kivelson and C.T. Russell (Eds.), Introduction to Space Physics (pp. 129-161). Cambridge: Cambridge University Press.

Callis, L. B., Baker D. N., Blake J. B., Lambeth J. D., Boughner R. E., Natarajan M., & Gorney D. J. (1991). Precipitating relativistic electrons: Their long-term effect on stratospheric odd nitrogen levels. Journal of Geophysical Research: Atmospheres, 96(D2), 2939–2976. https://doi.org/10.1029/90JD02184

Callis, L., M. Natarajan, D. Evans, and J. Lambeth, (1998a). Solar atmospheric coupling by electrons (SOLACE) 1. Effects of the May 12, 1997 solar event on the middle atmosphere. Journal of Geophysical Research: Atmospheres, 103(D21), 28405-28419. https://doi:10.1029/98JD02408

Callis, L., M. Natarajan, J. Lambeth, and D. Baker (1998b). Solar atmospheric coupling by electrons (SOLACE) 2. Calculated stratospheric effects of precipitating electrons, 1979-1988. Journal of Geophysical Research: Atmospheres, 103(D21), 28421-28438. https://doi:10.1029/98JD02407

111

Callis, L. B., M. Natarajan, and J. D. Lambeth, Solar atmospheric coupling by electrons (SOLACE), Part 3. Comparisons of simulations and observations, 1979–1997, issues and implications (2001). Journal of Geophysical Research: Atmospheres., 106, 7523

Cane, H. V., W. C. Erickson, and N. P. Prestage (2002). Solar flares, type III radio bursts, coronal mass ejections, and energetic particles, Journal of Geophysical Research, 107(A10), 1315. https://doi:10.1029/2001JA000320

Carlson, Jr., H.C. and A. Egeland (1995). The Aurora and the Auroral Ionosphere. In M.G. Kivelson and C.T. Russell (Eds.), Introduction to Space Physics (pp. 129-161). Cambridge: Cambridge University Press.

Chamberlin, P. C., Woods T. N., & Eparvier F. G. (2008). Flare Irradiance Spectral Model (FISM): Flare component algorithms and results. Space Weather, 6, S05001. https://doi.org/10.1029/2007SW000372

Chamberlin, P. C., Woods T. N., & Eparvier F. G. (2007). Flare Irradiance Spectral Model (FISM): Daily component algorithms and results. Space Weather, 5, S07005.

Charbonneau P., (2010). Dynamo Models of the Solar Cycle. Living Reviews in Solar Physics, 7, 3.

Clark D. H., Stephenson F. R. (1978). An interpretation of the pre-telescopic sunspot records from the orient. Q J R Astron Soc 19:387–410

Clilverd, M. A., C. J. Rodger, R. J. Gamble, T. Ulich, T. Raita, A. Seppälä, J. C. Green, N. R. Thomson, J.-A. Sauvaud, and M. Parrot (2010). Ground-based estimates of outer radiation belt energetic electron precipitation fluxes into the atmosphere. Journal of Geophysical Research, 115, A12304, doi:10.1029/2010JA015638

Clilverd, M. A., N. Cobbett, C. J. Rodger, J. B. Brundell, M. H. Denton, D. P. Hartley, … E. L. Spanswick (2013). Energetic electron precipitation characteristics observed from Antarctica during a flux dropout event. Journal of Geophysical Research: Space Physics, 118, 6921–6935. https;//doi:10.1002/2013JA019067

Codrescu, M., T. Fuller-Rowell, R. Roble, and D. Evans (1997). Medium energy particle precipitation influences on the mesosphere and lower thermosphere, Journal of Geophysical. Research: Space Physics, 102(A9), 19977-19987

Crutzen, P. J., Isaksen I. S. A., & Reid G. C. (1975). Solar proton events: Stratospheric sources of nitric oxide. Science, 189, 457–459. https://doi.org/10.1126/science.189.4201.457

Crutzen, P. J., (1979). The Role of NO and NO2 on the chemistry of the troposphere and stratosphere. Annual Review of Earth Planetary Science, 7, 443-472.

112

Cunnold, D. M., et al. (1991). Validation of SAGE II NO2 measurements. Journal of Geophysical Research, 96(D7), 12,913–12,925. doi:10.1029/91JD01344

Degenstein, D.A., Lloyd, N.D., Bourassa, A.E., Gattinger, R.L. & Llewellyn, E.J. (2005). Observations of mesospheric ozone depletion during the October 28, 2003 solar proton event by OSIRIS. Geophysical Research Letters, 32, L03S11. https://doi.org/10.1029/2004GL021521

Dennis, B. R., (1988). Solar flare hard X-ray observations. Solar Physics, 118, 49-94

Dennis, B. R., and D. M. Zarro (1993). The Neupert Effect: What can it tell us about the impulsive and gradual phases of solar flares? Solar Physics, 146, 177, doi:10.1007/BF00662178

Donnelly, R. F. (1976). Empirical models of solar flare x-ray and EUV emissions for use in studying their E and F region effects. Journal of Geophysical Research: Atmospheres, 81, 4745- 4753. https://doi.org/10.1029/JA081i025p04745

Eastman, T.E., E.W. Hones Jr., S.J. Bame, J.R. Asbridge (1976). The magnetospheric boundary layer: site of plasma, momentum, and energy transfer from the magnetosheath into the magnetosphere. Geophysical Research Letters, 3, 685

Eather, R., S. Mende, and R. Judge (1976). Plasma injection at synchronous orbit and spatial and temporal auroral morphology. Journal of Geophysical Research: Space Physics, 81(16), 2805- 2834. doi:10.1029/JA081i016p02805

Evans, D. S, and M. S. Greer (2000). Polar Orbiting Environmental Satellite Space Environmental Monitor-2: Instrument descriptions and archive data documentation, Tech. Mem., Natl. Atmos. And Oceanic Admin., Space Environ. Cent., Boulder, CO.

Evans, D. S., and M. S. Greer (2004). Polar orbiting environmental satellite space environment monitor-2: Instrument descriptions and archive data documentation, NOAA Tech. Mem. 1.4, Space Environ. Cent., Boulder, Colo.

Eyring, V., Arblaster J. M., Cionni I., Sedláček J., Perlwitz J., Young P. J., … Watanbe S. (2013). Long-term ozone changes and associated climate impacts in CMIP5 simulations. Journal of Geophysical Research: Atmospheres, 118, 5029–5060, https://doi.org/10.1002/jgrd.50316

Eyring, V., Bony, S., Meehl, G. A., Senior, C. A., Stevens, B., Stouffer, R.J., and Taylor, K. E. (2016). Overview of the Coupled Model Intercomparison Project Phase 6 (CMIP6) experimental design and organization. Geosciences Model Development, 9, 1937-1958. https://doi.org/10.5194/gmd-9-1937-2016

Fan, Y., (2009). Magnetic Fields in the Solar Convection Zone. Living Reviews in Solar Physics, 6, 4.

113

Fang, X., C. E. Randall, D. Lummerzheim, S. C. Solomon, M. J. Mills, D. R. Marsh, … G. Lu (2008). Electron impact ionization: A new 142 parameterization for 100 eV to 1 MeV electrons. Journal of Geophysical Research, 113, A09311, doi:10.1029/2008JA013384

Fang, X., C. E. Randall, D. Lummerzheim, W. Wang, G. Lu, S. C. Solomon, and R. A. Frahm (2010), Parameterization of monoenergetic electron impact ionization. Geophysical Research Letters, 37, L22106, doi:10.1029/2010GL045406

Fischer, H., et al. (2008). MIPAS: An instrument for atmospheric and climate research. Atmospheric Chemistry and Physics, 8, 2151-2188.

Fisk, L. A. and M. A. Lee (1980). Shock Acceleration of Energetic Particles in Corotating Interaction Regions in the Solar Wind. Astrophysical Journal, 237, 620-626.

Forbes, T., (2010). Models of coronal mass ejections and flares. In C. J. Schrijver and G. L. Siscoe (Eds.). Heliophysics: Space Storms and Radiation: Causes and Effects (pp. 159-193). Cambridge: Cambridge University Press.

Frahm, R. A., J. D. Winningham, J. R. Sharber, R. Link, G. Crowley, E. E. Gains, … T. A. Potemra (1997). The diffuse aurora: A significant source of ionization in the middle atmosphere. Journal of Geophysical Research, 102, 28,203

Fuller-Rowell, T. J., and S. C. Solomon (2010). Flares, coronal mass ejections, and atmospheric responses. In C. J. Schrijver and G. L. Siscoe (Eds.), Heliophsyics: Space Storms and Radiation: Causes and Effects. (pp. 321–357). Cambridge: Cambridge University Press.

Funke, B., M. Lo´pez-Puertas, S. Gil-Lo´pez, T. von Clarmann, G. P. Stiller, H. Fischer, and S. Kellmann (2005). Downward transport of upper atmospheric NOx into the polar stratosphere and lower mesosphere during the Antarctic 2003 and Arctic 2002/2003 winters. Journal of Geophysical Research, 110, D24308, doi:10.1029/2005JD006463

Funke, B., Puertas, M. L., Holt, L., Randall, C. E., Stiller, G. P., and von Clarmann, T., Hemispheric distributions and interannual variability of NOy produced by energetic particle precipitation in 2002–2012. Journal of Geophysical Research, 119, 13565–13582, doi:10.1002/2014JD022423, 2014b

Funke, B., Baumgaertner A., Calisto M., Egorova T., Jackman C. H., Kieser J., … Wissing J. M. (2011). Composition changes after the "Halloween" solar proton event: the High Energy Particle Precipitation in the Atmosphere (HEPPA) model versus MIPAS data intercomparison study. Atmospheric Chemistry and Physics, 11, 9089-9139. https://doi.org/10.5194/acp-11-9089-2011

Garcia, R. and Solomon, S. (1983). A numerical model of the zonally averaged dynamical and chemical structure of the middle atmosphere. Journal of Geophysical Research. 88. 1379-1400. 10.1029/JC088iC02p01379

114

Garcia, R. R., Smith, A. K., Kinnison, D. E., de la C_amara, _A., & Murphy, D. J. (2017). Modification of the gravity wave parameterization in the Whole Atmosphere Community Climate Model: Motivation and results. Journal of the Atmospheric Sciences, 74, 275–291. https://doi.org/ 10.1175/JAS-D-16-0104.1

Gent, P. R., Danabasoglu, G., Donner, L. J., Holland, M. M., Hunke, E. C., Jayne, S. R., … Zhang, M. (2011). The Community Climate System Model Version 4. Journal of Climate, 24, 4973–4991. https://doi.org/10.1175/2011JCLI4083.1

Goody, R.M. and Yung, Y.L. (1989). Atmospheric radiation: Theoretical basis. Oxford University Press, New York/Oxford

Gopalswamy, N., Yashiro S., & Akiyama S., (2006). Coronal mass ejections and space weather due to extreme events. Paper presented at ILWS Workshop, Goa, India.

Gopalswamy, N., S. Yashiro, and S. Akiyama (2007), Geoeffectiveness of halo coronal mass ejections, Journal of Geophysical Research, 112, A06112. https://doi:10.1029/2006JA012149

Gordley, L. L., et al. (1996). Validation of nitric oxide and nitrogen dioxide measurements made by the Halogen Occultation Experiment for UARS platform. Journal of Geophysical Research, 101, 10,241– 10,266, doi:10.1029/95JD02143

Green J. (2010). Energization of trapped particles. In C. J. Schrijver and G. L. Siscoe (Eds.), Heliophysics: Space Storms and Radiation: Causes and Effects (pp. 293-321). Cambridge: Cambridge University Press.

Green, J. C. (2013). MEPED Telescope Data Processing Algorithm Theoretical Basis Document, NOAA NGDC.

Gu, X., Z. Zhao, B. Ni, Y. Shprits, and C. Zhou (2011). Statistical analysis of pitch angle distribution of radiation belt energetic electrons near the geostationary orbit: CRRES observations. Journal of Geophysical Research, 116, A01208, doi:10.1029/2010JA016052

Hale, G. E., Ellerman, F., Nicholson, S. B., and Joy, A. H. (1919). The Magnetic Polarity of Sun- spots. Astrophysical Journal, 49, 153.

Harvey, V. L., C. E. Randall, and R. L. Collins (2015). Chemical definition of the mesospheric polar vortex. Journal of Geophysical Research: Atmospheres, 120, 10,166–10,179, doi:10.1002/ 2015JD023488

Hendrickx, K., Megner, L., Marsh, D. R., Gumbel, J., Strandberg, R., & Martinsson, F. (2017). Relative importance of nitric oxide physical drivers in the lower thermosphere. Geophysical Research Letters, 44, 10,081–10,087. https://doi.org/10.1002/2017GL074786

Holt, L. A., C. E. Randall, E. D. Peck, D. R. Marsh, A. K. Smith, and V. L. Harvey (2013). The influence of major sudden stratospheric warmings and elevated stratopause events on the effects

115 of energetic particle precipitation in WACCM. Journal of Geophysical Research: Atmospheres, 118, 11,636-11,646. https:\\doi:10.1002/2013JD020294

Holton, J. R., (2004). Dynamic Meteorology. Elsevier Academic Press, San Diego, California.

Hudson, H. (2010). Observations of solar and stellar eruptions, flares, and jets. In C. J. Schrijver and G. L. Siscoe (Eds.), Heliophysics: Space Storms and Radiation: Causes and Effects (pp. 123-159). Cambridge: Cambridge University Press.

Hurrell, J. W., Holland M. M., Gents P. R., Ghan S., Kay J. E., Kushner P. J., … Marshall S. (2013). The Community Earth System Model: A framework for collaborative research. Bulletin of the American Meteorological Society, 94, 1339–1360. https://doi.org/10.1175/BAMS-D-12- 00121.1

Jackman, C. H., Frederick J. E., & Stolarski, R.S. (1980). Production of Odd Nitrogen in the Stratosphere and Mesosphere: An Intercomparison of Source Strengths. Journal of Geophysical Research: Atmospheres, 85, 7495-7505. https://doi.org/10.1029/JC085iC12p07495

Jackman, C. H., Cerniglia, M. C., Nielsen, E. J., Allen, D. J., Zawodny, J. M., McPeters, R. D., … Rood, R. B. (1995). Two-dimensional and three-dimensional model simulations, measurements and interpretation of the influence of the October 1989 solar proton events on the middle atmosphere. Journal of Geophysical Research: Atmospheres, 100, 11641– 11660. https://doi.org/10.1029/95JD00369

Jackman, C. H. and McPeters, R. D.: The Effect of Solar Proton Events on Ozone and Other Constituents, in: Solar Variability and its Effects on Climate, Geophys. Mon, 141, 305–319, 2004.

Jackman, C. H., DeLand, M. T., Labow, G. J., Fleming, E. L., Weisenstein, D. K., Ko, M., … Russell, J. M. (2005). The influence of the several very large solar proton events in years 2000 – 2003 on the neutral middle atmosphere. Advances in Space Research, 35, 445–450. https://doi.org/10.1016/j.asr.2004.09.006

Jackman, C. H., M. T. DeLand, G. J. Labow, E. L. Fleming, and M. Lopez-Puertas (2006). Satellite measurements of middle atmospheric impacts by solar proton events in solar cycle 23. Space Science Reviews, 125, 381-391. https://doi.org/10.1007/s11214-006-9071-4

Jackman, C. H., Marsh, D. R., Vitt, F. M., Garcia, R. R., Fleming, E. L., Labow, G. J., … Stiller, G. P. (2008). Short and medium-term atmospheric constituent effects of very large solar proton events. Atmospheric Chemistry and Physics, 8, 765-785. https://doi.org/10.5194/acp-8-765-2008

Jackman, C. H., Marsh, D. R., Vitt, F. M., Garcia, R. R., Randall, C. E., Fleming, E. L., & Frith, S. M. (2009). Long-term middle atmospheric influence of very large solar proton events. Journal of Geophysical Research: Atmospheres, 114, D11304. https://doi.org/10.1029/2008JD011415

116

Jackman, C. H., Randall, C. E., Harvey, V. L., Wang, S., Fleming, E. L., Lopez-Puertas, M., Funke, B., & Bernath, P. F. (2014). Middle atmospheric changes caused by the January and March 2012 solar proton events. Atmospheric Chemistry and Physics, 14, 1025-1038. https://doi.org/10.5194/acp-14-1025-2014

Jin, J. J., et al. (2005). Co-located ACE-FTS and Odin/SMR stratospheric-mesospheric CO 2004 measurements and comparison with a GCM. Geophysical Research Letters, 32, L15S03, doi:10.1029/2005GL022433

Jordanova, V. K., J. Albert, and Y. Miyoshi (2008). Relativistic electron precipitation by EMIC waves from self-consistent global simulations. Journal of Geophysical Research: Space Physics, 113, A00A10, doi:10.1029/2008JA013239

Kerzenmacher, T., Wolff, M. A., Strong, K., Dupuy, E., Walker, K. A., Amekudzi, L. K., … Zou, J. (2008). Validation of NO2 and NO from the Atmospheric Chemistry Experiment (ACE). Atmospheric Chemistry and Physics, 8, 5801-5841. https://doi.org/10.5194/acp-8-5801-2008

Kinnison, D. E., Brasseur, G. P., Walters, S., Garcia, R. R., Marsh, D. R., Sassi, F., … Simmons, A. J. (2007). Sensitivity of chemical tracers to meteorological parameters in the MOZART-3 chemical transport model. Journal of Geophysical Research: Atmospheres, 112, D20302. https://doi.org/10.1029/2006JD007879

Krauss-Varban D. (2010). Particle acceleration in shocks. In C. J. Schrijver and G. L. Siscoe (Eds.), Heliophysics: Space Storms and Radiation: Causes and Effects (pp. 209-233). Cambridge: Cambridge University Press.

Lam, M. M., R. B. Horne, N. P. Meredith, S. A. Glauert, T. Moffat‐Griffin, and J. C. Green (2010). Origin of energetic electron precipitation >30 keV into the atmosphere. Journal of Geophysical Research: Space Physics, 115, A00F08, doi:10.1029/2009JA014619

Lamarque, J.-F., Emmons, L. K., Hess, P. G., Kinnison, D. E., Tilmes, S., Vitt, F., Heald, C. L., Holland, E. A., Lauritzen, P. H., Neu, J., Orlando, J. J., Rasch, P. J., and Tyndall, G. K.: CAM- chem: description and evaluation of interactive atmospheric chemistry in the Community Earth System Model, Geosciences Model Development, 5, 369-411. https://doi.org/10.5194/gmd-5- 369-2012, 2012

Lazarev, V. I. (1967). Absorption of the energy of an electron beam in the upper atmosphere. Geomagnetism and Aeronomy, 7, 219.

Lei, J., Thayer, J.P., Forbes, J.M., Sutton, E.K., Nerem, R.S., … Veronig, A.M. (2008c). Global thermospheric density variations caused by high-speed solar wind streams during the declining phase of solar cycle 23. Journal of Geophysical Research: Atmospheres, 113, A11303, doi:10.1029/2008JA013433

117

Lei, J., Thayer, J.P., Wang, W., McPherron, R.L., (2011). Impact of CIR Storms on Thermosphere Density Variability during the Solar Minimum of 2008. Solar Physics, 274, 427- 437, doi: 10.1007/s11207-010-9563-y

Lean, J., J. Beer, and R. Bradley, Reconstruction of solar irradiance since 1610: Implications for climate change.Geophysical Research Letters, 22, 3195-3198, 1995a

Liu, H.-L., Bardeen, C. G., Foster, B. T., Lauritzen, P., Liu, J., Lu, G., . . . Wang, W. (2018). Development and validation of the Whole Atmosphere Community Climate Model with thermosphere and ionosphere extension (WACCM-X 2.0). Journal of Advances in Modeling Earth Systems, 10, 381–402. https://doi.org/ 10.1002/2017MS001232

Liu, J. Y., Lin, C. H., Chen, Y. I., Lin, Y. C., Fang, T. W., Chen, C. H., Chen, Y. C., & Hwang, J. J. (2006). Solar flare signatures of the ionospheric GPS total electron content. Journal of Geophysical Research: Space Physics, 111, A05308. https://doi.org/10.1029/2005JA011306

López-Puertas, M., B. Funke, S. Gil-López, T. von Clarmann, G. P. Stiller, M. Ho¨pfner, … C. H. Jackman (2005). Observation of NOx enhancement and ozone depletion in the Northern and Southern Hemispheres after the October–November 2003 solar proton events, Journal of Geophysical Research: Space Physics, 110, A09S43. https://doi:10.1029/2005JA011050

López-Puertas, M., Funke, B., Gil-Lopez, S., von Clarmann, T., Stiller, G. P., Höpfner, M., … Jackman, C. H. (2005). HNO3, N2O5, and ClONO2 enhancements after the October–November 2003 solar proton events. Journal of Geophysical Research: Space Physics, 110, A09S44. https://doi.org/10.1029/2005JA011051

Luhmann J.G. (1995). Plasma Interactions with Unmagnetized Bodies. In M.G. Kivelson and C.T. Russell (Eds.), Introduction to Space Physics (pp. 129-161). Cambridge: Cambridge University Press.

Lummerzheim, D. (1992). Comparison of energy dissipation functions for high energy auroral electrons and ion precipitation. Rep. UAG-R-318, Geophys. Inst., Univ. of Alaska-Fairbanks, Fairbanks.

Manney, G. L., and J. L. Sabutis (2000). Development of the polar vortex in the 1999 – 2000 Arctic winter stratosphere. Geophysical Research Letters, 27, 2589–2592.

Manney, G. L., K. Krüger, J. L. Sabutis, S. A. Sena, and S. Pawson (2005). The remarkable 2003–2004 winter and other recent warm winters in the Arctic stratosphere since the late 1990s. Journal of Geophysical Research: Atmospheres, 110, D04107, doi:10.1029/2004JD005367.

Manney, G. L., et al. (2008). The evolution of the stratopause during the 2006 major warming: Satellite data and assimilated meteorological analyses. Journal of Geophysical Research: Atmospheres., 113, D11115, doi:10.1029/2007JD009097.

118

Manney, G. L., M. J. Schwartz, K. Krüger, M. L. Santee, S. Pawson, J. N. Lee, W. H. Daffer, R. A. Fuller, and N. J. Livesey (2009a). Aura Microwave Limb Sounder observations of dynamics and transport during the record-breaking 2009 Arctic stratospheric major warming, Geophysical Research Letters, 36, L12815. https://doi:10.1029/2009GL038586

Manney, G. L., et al. (2009b). Satellite observations and modeling of transport in the upper troposphere through the lower mesosphere during the 2006 major stratospheric sudden warming. Atmospheric Chemistry and Physics, 9, 4775–4795

Marsh, D. R., Mills, M. J., Kinnison, D. E., Lamarque, J. F., Calvo, N., & Polvani, L. M. (2013). Climate change from 1850 to 2005 simulated in CESM1(WACCM). Journal of Climate, 26, 7372–7391. https://doi.org/10.1175/JCLI-D-12-00558.1

Matthes, K., Funke, B., Andersson, M. E., Barnard, L., Beer, J., Charbonneau, P., … Versick, S. (2017). Solar forcing for CMIP6 (v3.2). Geoscientific Model Development, 10, 2247-2302. https://doi.org/10.5194/gmd-10-2247-2017

Meredith, N. P., R. B. Horne, M. M. Lam, M. H. Denton, J. E. Borovsky, and J. C. Green (2011). Energetic electron precipitation during high‐speed solar wind stream driven storms. Journal of Geophysical Research: Space Physics, 116, A05223, doi:10.1029/2010JA016293.

Millan, R. M., and R. M. Thorne (2007). Review of radiation belt relativistic electron loss. Journal of Atmospheric and Solar-Terrestrial Physics, 69, 362–377, doi:10.1016/j.jastp.2006.06.019

Miyoshi, Y., & Kataoka, R. ( 2005). Ring current ions and radiation belt electrons during geomagnetic storms driven by coronal mass ejections and corotating interaction regions. Geophysical Research Letters, 32, L21105. https://doi.org/10.1029/2005GL024590

Moreno-Villanueva M., M. Wong, T. Lu, Y. Zhang, H. Wu (2017). Interplay of space radiation and microgravity in DNA damage and DNA damage response. Nature, 3, 14. https://doi.org/10.1038/s41526-017-0019-7

Nash, E., P. Newman, J. Rosenfield, and M. Schoeberl (1996). An objective determination of the polar vortex using Ertel’s potential vorticity. Journal of Geophysical Research: Atmospheres, 101(D5), 9471–9478, doi:10.1029/96JD00066

Nesse Tyssøy, H., M. I. Sandanger, L.-K. G. Ødegaard, J. Stadsnes, & A. Aasnes1 A. E. (2016). Energetic Electron Precipitation into the Middle Atmosphere - Constructing the Loss Cone Fluxes from MEPED POES. Journal of Geophysical Research: Space Physics, 121, 5693-5707. doi: 10.1002/2016JA022752

Neupert, W. N. (1968). Comparison of solar X-ray line emission with microwave emission during flares. Journal of Astrophysics, 153, L59, doi:10.1086/180220

119

Newnham, D. A., P. J. Espy, M. A. Clilverd, C. J. Rodger, A. Seppälä, D. J. Maxfield, P. Hartogh, K. Holmén, and R. B. Horne (2011). Direct observations of nitric oxide produced by energetic electron precipitation into the Antarctic middle atmosphere. Geophysical Research Letters: Space Physics, 38, L20104, doi:10.1029/2011GL048666

Newnham, D. A., Clilverd, M. A., Rodger, C. J., Hendrickx, K., Megner, L., Kavanagh, A. J., et al. (2018). Observations and modeling of increased nitric oxide in the Antarctic polar middle atmosphere associated with geomagnetic storm-driven energetic electron precipitation. Journal of Geophysical Research: Space Physics, 123, 6009–6025. https://doi.org/ 10.1029/2018JA025507

O’Brien, T. P., and S. K. Morley (2011). Documentation of C Inversion Library, from the IRBEM library,

Orsolini, Y., M. L. Santee, G. L. Manney, and C. E. Randall (2005). An upper stratospheric layer of enhanced HNO3 following exceptional solar storms. Geophysical Research Letters, 32(12), L12S01, doi:10.1029/2004/GL021588

Orsolini, Y. J., Smith‐Johnsen, C., Marsh, D. R., Stordal, F., Rodger, C. J., Verronen, P. T., & Clilverd, M. A. (2018). Mesospheric nitric acid enhancements during energetic electron precipitation events simulated by WACCM‐D. Journal of Geophysical Research: Atmospheres, 123, 6984– 6998

Pedatella, N. M., K. Raeder, J. L. Anderson, and H.-L. Liu (2014). Ensemble data assimilation in the Whole Atmosphere Community Climate Model. Journal of Geophysical Research: Atmospheres, 119, 9793–9809. doi:10.1002/2014JD021776

Parsons J. L., and L.W. Townsend (2000). Interplanetary Crew Dose Rates for the August 1972 Solar Particle Event. Radiation Research, 153, 729-733. https://doi.org/10.1667/0033- 7587(2000)153[0729:ICDRFT]2.0.CO;2

Peck, E. D., Randall, C. E., Green, J. C., Rodriguez, J. V., & Rodger, C. J. (2015). POES MEPED differential flux retrievals and electron channel contamination correction. Journal of Geophysical Research: Space Physics, 120, 4596–4612. https://doi.org/10.1002/2014JA020817

Pettit, J., Randall, C. E., Marsh, D. R., Bardeen, C. G., Qian, L., Jackman, C. H., et al. (2018). Effects of the September 2005 solar flares and solar proton events on the middle atmosphere in WACCM. Journal of Geophysical Research: Space Physics, 123, 5747–5763. https://doi.org/10.1029/2018JA025294

Picone, J. M., Hedin, A.E., Drob, D.P., & Aikin, A.C. (2002). NRLMSISE-00 empirical model of the atmosphere: Statistical comparisons and scientific issues. Journal of Geophysical Research, 107(A12), 1468. https://doi:10.1029/2002JA009430, 2002

120

Porter, H. S., Jackman, C. H., & Green, A. E. S. (1976). Efficiencies for production of atomic nitrogen and oxygen by relativistic proton impact in air. Journal of Chemical Physics, 65, 154 – 167. https://doi.org/10.1063/1.432812

Priest E.R. (1995). The Sun and its Magnetohydrodynamics. In M.G. Kivelson and C.T. Russell (Eds.), Introduction to Space Physics (pp. 129-161). Cambridge: Cambridge University Press.

Qian, L., Burns, A. G., Chamberlin, P. C., & Solomon, S. C. (2010). Flare location on the solar disk: Modeling the thermosphere and ionosphere response. Journal of Geophysical Research, 115, A09311. https://doi.org/10.1029/2009JA015225

Qian, L., Burns, A. G., Chamberlin, P. C., & Solomon, S. C. (2011). Variability of thermosphere and ionosphere responses to solar flares. Journal of Geophysical Research: Space Physics, 116, A10309, https://doi.org/10.1029/2011JA016777

Randall, C.E., Bevilacqua, R. M., Rusch, D. W., Hoppel, K. W., & Lumpe, J. D. (1998). POAM II measurements of NO2, 1993-1996. Journal of Geophysical Research: Atmospheres, 103, 28,361-28,371. https://doi.org/10.1029/98JD02092

Randall, C. E., Siskind, D. E., & Bevilacqua, R. M. (2001). Stratospheric NOx enhancements in the southern hemisphere vortex in winter/spring of 2000. Geophysical Research Letters, 28 (12), 2385-2388. https://doi.org/10.1029/2000GL012746

Randall, C. E., et al., Validation of POAM III NO2 measurements. Journal of Geophysical Research, 107(D20), 4432. https://doi:10.1029/2001JD001520

Randall, C.E., Harvey, V. L., Manney, G. L, Orsolini, Y., Codrescu, M., Sioris, C., … Russell, J. M. III (2005). Stratospheric effects of energetic particle precipitation in 2003-2004. Geophysical Research Letters. 32, L05802. https://doi.org/10.1029/2004GL02203

Randall, C. E., V. L. Harvey, C. S. Singleton, P. F. Bernath, C. D. Boone, and J. U. Kozyra (2006). Enhanced NOx in 2006 linked to strong upper stratospheric Arctic vortex, Geophysical Research Letters, 33, L18811. https://doi:10.1029/2006GL027160

Randall, C. E., V. L. Harvey, C. S. Singleton, S. M. Bailey, P. F. Bernath, M. Codrescu, ... J. M. Russell III (2007). Energetic particle precipitation effects on the Southern Hemisphere stratosphere in 1992–2005. Journal of Geophysical Research: Atmospheres, 112, D08308, https://doi:10.1029/2006JD007696

Randall, C. E., V. L. Harvey, D. E. Siskind, J. France, P. F. Bernath, C. D. Boone, and K. A. Walker (2009). NOx descent in the Arctic middle atmosphere in early 2009. Geophysical Research Letters, 36, L18811, doi:10.1029/2009GL039706

Randall, C. E., Harvey, V. L., Holt, L. A., Marsh, D. R., Kinnison, D., Funke, B., & Bernath, P. F. (2015). Simulation of energetic particle precipitation effects during the 2003–2004 Arctic winter. Journal of Geophysical Research: Space Physics, 120, 5035-5048.

121 https://doi.org/10.1002/2015JA021196

Rault, D. F. (2004).Ozone, NO2 and aerosol retrieval from SAGE III limb scatter measurements, in Proc. SPIE 5571. Remote Sensing of Clouds and the Atmosphere IX, 205, doi:10.1117/12.564899

Richardson, I. G. (2003). Energetic particles and corotating interaction region in the solar wind. Space Science Reviews, 111, 267– 376

Richardson, I. G., et al. (2006). Major geomagnetic storms (Dst < -100 nT) generated by corotating interaction regions. Journal of Geophysical Research, 111, A07S09, doi:10.1029/2005JA011476

Rienecker M. M, Suarez, M. J., Gelaro, R., Todling, R., Bacmeister, J., Liu, E., … Woollen J. (2011). MERRA: NASA’s modern-era retrospective analysis for research and applications. Journal of Climate, 24, 3624–3648. https://doi.org/10.1175/JCLI-D-11-00015.1

Roble, R. G., and Ridley, E. C. (1987). An auroral model for the NCAR thermospheric general circulation model (TGCM). Annales Geophysicae, 5A(6), 369-382

Roble, R. G., and Rees M. H., (1977). Time-dependent studies of the aurora: Effects of particle precipitation on the dynamic morphology of ionospheric and atmospheric properties. Planetary Space Sciences, 22, 623-631.

Rodger, C. J., T. Raita, M. A. Clilverd, A. Seppälä, S. Dietrich, N. R. Thomson, and T. Ulich (2008) Observations of relativistic electron precipitation from the radiation belts driven by EMIC waves. Geophysical Research Letters, 35, L16106, doi:10.1029/2008GL034804

Rodger, C. J., M. A. Clilverd, J. C. Green, and M. M. Lam (2010a). Use of POES SEM-2 observations to examine radiation belt dynamics and energetic electron precipitation into the atmosphere. Journal of Geophysical Research: Space Physics, 115, A04202, doi:10.1029/2008JA014023

Rodger, C. J., A. J. Kavanagh, M. A. Clilverd, and S. R. Marple (2013). Comparison between POES energetic electron precipitation observations and riometer absorptions: Implications for determining true precipitation fluxes. Journal of Geophysical Research: Space Physics, 118, 7810–7821, doi:10.1002/2013JA019439

Rohen, G., von Savigny, C., Sinnhuber, M., Llewellyn, E. J., Kaiser, J. W., Jackman, C. H. (2005). Ozone depletion during the solar proton events of October/November 2003 as seen by SCIAMACHY. Journal of Geophysical Research: Space Physics, 110, A09S39. https://doi.org/10.1029/2004JA010984

Rottman, G.J., Woods T. N., & Sparn, T. P. (1993). Solar Stellar Irradiance Comparison Experiment I: 1 instrument design and operation. Journal of Geophysical Research: Atmospheres, 98, 10667–10677. https://doi.org/10.1029/93JD00462

122

Rozanov, E., L. Callis, M. Schlesinger, F. Yang, N. Andronova, and V. Zubov (2005). Atmospheric response to NOy source due to energetic electron precipitation. Geophysical Research Letters, 32, L14811. doi:10.1029/ 2005GL023041

Rozanov, E., Callisto, M., Egorova, T., Peter, T. & Schmutz, W. (2012). The influence of precipitating energetic particles on atmospheric chemistry and climate. Surv. Geophys. 33, 483–501.

Rusch, D., J. Gerard, S. Solomon, P. Crutzen, and G. Reid (1981). The effect of particle precipitation events on the neutral and ion chemistry of the middle atmosphere -I. Odd nitrogen, Planetary Space Sciences,,29(7), 767-774. doi:10.1016/0032-0633(81)90048-9

Russell, J. M. III, L. L. Gordley, J. H. Park, S. R. Drayson, W. D. Hesketh, R. J. Cicerone, A. F. Tuck, J. E. Frederick, J. E. Harries, and P. J. Crutzen (1993). The Halogen Occultation 152 Experiment. Journal of Geophysical Research, 98, 10,777-10,797

Seppälä, A., Clilverd, M.A. & Rodger, C.J. (2007). NOx enhancements in the middle atmosphere during 2003–2004 polar winter: Relative significance of solar proton events and the aurora as a source. Journal of Geophysical Research: Atmospheres, 112, D23303. https://doi.org/10.1029/2006JD008326

Seppälä, A., Verronen, P. T., Kyrölä, E., Hassinen, S., Backman, L., Hauchecorne, A., Bertaux, J. L., & Fussen, D. (2004). Solar proton events of October –November 2003: Ozone depletion in the Northern Hemisphere polar winter as seen by GOMOS/Envisat. Geophysical Research Letters, 31, L19107, https://doi.org/10.1029/2004GL021042

Seppälä, A., C. E. Randall, M. A. Clilverd, E. Rozanov, and C. J. Rodger (2009). Geomagnetic activity and polar surface air temperature variability, Journal of Geophysical Research, 114, A10312. doi:10.1029/2008JA014029

Seppälä, A., Douma, E., Rodger, C. J., Verronen, P. T., Clilverd, M. A., & Bortnik, J. (2018). Relativistic electron microburst events: Modeling the atmospheric impact. Geophysical Research Letters, 45, 1141–1147. https://doi.org/10.1002/2017GL075949

Sheese, P. E., Walker, K. A., Boone, C. D., McLinden, C. A., Bernath, P. F., Bourassa, A. E., … Zawodny, J. M. (2016). Validation of ACE-FTS version 3.5 NOy species profiles using correlative satellite measurements. Atmospheric Measurement Techniques, 9, 5781-5810. https://doi.org/10.5194/amt-9-5781-2016

Shibata, K. & Magara,T. (2011). Solar Flares: Magnetohydrodynamic Processes. Living Reviews in Solar Physics, 8, 6. https://doi.org/10.12942/lrsp-2011-6

Siskind, D. E., S. D. Eckermann, L. Coy, J. P. McCormack, and C. E.Randall (2007). On recent interannual variability of the Arctic winter mesosphere: Implications for tracer descent. Geophysical Research Letters, 34, L09806. doi:10.1029/2007GL029293

123

Siskind, D. E., Eckermann, S. D., McCormack, J. P., Coy, L., Hoppel, K. W., and Baker, N. L. (2010). Case studies of the mesospheric response to minor, major and extended stratospheric warmings. Journal of Geophysical Research, 115, D00N03. htttps://doi:10.1029/2010JD014114, 2010

Smith-Johnsen, C., Marsh, D. R., Orsolini, Y., Nesse Tyssøy, H., Hendrickx, K., Sandanger, M. I., et al. (2018). Nitric oxide response to the April 2010 electron precipitation event: Using WACCM and WACCM-D with and without medium-energy electrons. Journal of Geophysical Research: Space Physics, 123, 5232–5245. https://doi.org/10.1029/2018JA025418

Solomon, S., Rusch, D. W., Gérard, J. C., Reid, G. C., & Crutzen, P. J. (1981). The effect of particle precipitation events on the neutral and ion chemistry of the middle atmosphere: II. Odd hydrogen. Planetary and Space Science, 8, 885–893. https://doi.org/10.1016/0032- 0633(81)90078-7

Solomon, S., Crutzen, P. J, & Roble, R. G. (1982). Photochemical coupling between the thermosphere and the lower atmosphere: 1. Odd nitrogen from 50 to 120 km. Journal of Geophysical Research: Oceans, 87(C9), 7206–7220. https://doi.org/10.1029/JC087iC09p07206

Solomon, S. C., C. A. Barth, and S. M. Bailey (1999). Auroral production of nitric oxide measured by the Student Nitric Oxide Explorer. Geophysical Research Letters, 26, 1259

Solomon, S. C., & Qian, L. (2005). Solar extreme-ultraviolet irradiance for general circulation models. Journal of Geophysical Research: Space Physics, 110, A10306. https://doi.org/10.1029/2005JA011160

Solomon, S., Kinnison, D., Bandoro, J., & Garcia, R. (2015). Simulation of polar ozone depletion: An update. Journal of Geophysical Research: Atmospheres, 120, 7958–7974. https://doi.org/10.1002/2015JD023365

Solomon, S. C., Liu, H.-L., Marsh, D. R.,McInerney, J. M., Qian, L., & Vitt, F. M.(2018). Whole atmosphere simulationof anthropogenic climate change. Geophysical Research Letters, 45,1567– 1576. https://doi.org/10.1002/2017GL076950

Spencer, L. V. (1959). Energy Dissipation by Fast Electrons. NBS Monograph, 1, 70 pp.

Summers, D., B. Ni, and N. P. Meredith (2007). Timescales for radiation belt electron acceleration and loss due to resonant wave-particle interactions: 2. Evaluation for VLF chorus, ELF hiss, and electromagnetic ion cyclotron waves. Journal of Geophysical Research, 112, A04207. doi:10.1029/2006JA011993

Thorne, R. M. (1980). The importance of energetic particle precipitation on the chemical composition of the middle atmosphere, Pure and Applied Geophysics, 118, 128–151

124

Thurairajah, B., R. L. Collins, V. L. Harvey, R. S. Lieberman, and K. Mizutani (2010). Rayleigh lidar observations of reduced gravity wave activity during the formation of an elevated stratopause in 2004 at Chatanika, Alaska (65°N, 147°W), Journal of Geophysical Research, 115, D13109, doi:10.1029/2009JD013036

Tomczyk, S., McIntosh, S. W., Keil, S. L., Judge, P. G., Schad, T., Seeley, D. H., & Edmondson, J. (2007). Science, 317, 1192. https://doi.org/ 10.1126/science.1143304

H. Nesse Tyssøy, M. I. Sandanger, L.-K. G. Ødegaard, J. Stadsnes, & A. Aasnes1 A. E. (2016). Energetic Electron Precipitation into the Middle Atmosphere - Constructing the Loss Cone Fluxes from MEPED POES. Journal of Geophysical Research: Space Physics, 121, 5693-5707. doi: 10.1002/2016JA022752

Tsurutani, B. T., W. D. Gonzalez, A. L. C. Gonzalez, F. Tang, J. K. Arballo,and M. Okada (1995). Interplanetary origin of geomagnetic activity in thedeclining phase of the solar cycle, Journal of Geophysical Research, 100, 21,717

Tsurutani, B. T., Judge, D. L., Guarnieri, F. L., Gangopadhyay, P., Jones, A. R., Nuttall, J., … Viereck, R. (2005). The October 28, 2003 extreme EUV solar flare and resultant extreme ionospheric effects: Comparison to other Halloween events and the Bastille Day event. Geophysical Research Letters, 32, L03S09. https://doi.org/10.1029/2004GL021475

Tsurutani, B. T., Verkhoglyadova. O. P., Mannucci, A. J., Lakhina., G. S., Li. G., & Zank, G. P. (2009). A brief review of ‘‘solar flare effects’’ on the ionosphere. Radio Science, 44, RS0A17, https://doi.org/10.1029/2008RS004029

Turner, D. L., Y. Y. Shprits, M. Hartinger, and V. Angelopoulos (2012b). Explaining sudden losses of relativistic electrons during geomagnetic storms. Nature Physics, 8, doi:10.1038/NPHYS2185

Turner, D. L., V. Angelopoulos, W. Li, J. Bortnik, B. Ni, Q.Ma, … J. V. Rodriguez (2014). Competing source and loss mechanisms due to wave-particle interactions in Earth’s outer radiation belt during the 30 September to 3 October 2012 geomagnetic storm. Journal of Geophysical Research: Space Physics, 119, 1960–1979, doi:10.1002/2014JA019770

Turunen, E., Verronen, P. T., Seppälä, A., Rodger, C. J., Clilverd, M. A., Tamminen, J., … Ulich, T. (2009). Impact of different energies of precipitating particles on NOx generation in the middle and upper atmosphere during geomagnetic storms. Journal of Atmospheric and Solar- Terrestrial Physics, 71, 1176–1189. https://doi.org/10.1016/j.jastp.2008.07.005

Vampola, A. L. (1998), Outer zone energetic electron environment update, in Conference on the High Energy Radiation Background in Space, 128–136, IEEE Press, Piscataway, N. J. van de Kamp, M., A. Seppälä, M. A. Clilverd, C. J. Rodger, P. T. Verronen, and I. C. Whittaker

125

(2016). A model providing long-term data sets of energetic electron precipitation during geomagneticStorms. Journal of Geophysical Research: Atmospheres, 121, 12,520–12,540. doi:10.1002/2015JD024212

Verronen, P. T., M. E. Andersson, A. Kero, C.-F. Enell, J. M. Wissing, E. R. Talaat, K. Kauristie, M. Palmroth, T. E. Sarris, and E. Armandillo (2015). Contribution of proton and electron precipitation to the observed electron concentration in October-November 2003 and September

Verronen, P. T., M. E. Andersson, D. R. Marsh, T. Kovács, and J. M. C. Plane (2016). WACCM-D—Whole Atmosphere Community Climate Model with D-region ion chemistry. Journal of Advances in Modeling Earth Systems , 8, 954–975, doi:10.1002/2015MS000592

Vitt, F. M., & Jackman C. H. (1995). A Comparison of Sources of Odd Nitrogen Production from 1974 through 1993 in the Earth's Atmosphere as Calculated Using a Two-Dimensional Model. Journal of Geophysical Research: Atmospheres, 101, 6729-6739. https://doi.org/10.1029/95JD03386

Vršnak B., Temmer M., Veronig A.M., (2007). Coronal Holes and SolarWind High-Speed Streams: II. Forecasting the Geomagnetic Effects. Solar Physics, 240, 331-346. doi: 10.1007/s11207-007-0311-x

Wagner, W. J. (1984). Coronal mass ejections. Annual Review of Astronomy and Astrophysics, 22, 267.

Wittmann, A. D. and Xu, Z. T., (1987). A catalogue of sunspot observations from 165 BC to AD 1684. Astronomy and Astrophysics Supplement Series, 70, 1, 83-94.

Whittaker, I. C., C. J. Rodger, M. A. Clilverd, and J.-A. Sauvaud (2014). The effects and correction of the geometric factor for the POES/MEPED electron flux instrument using a multi- satellite comparison. Journal of Geophysical Research: Space Physics, 119, 6386–6404. https://doi.org/10.1002/2014JA020021.

Wissing, J. M., and M.-B. Kallenrode (2009). Atmospheric Ionization Module Osnabruck (AIMOS): A 3-D model to determine atmospheric ionization by energetic charged particles from different populations. Journal of Geophysical Research: Space Physics, 114, A06104, doi:10.1029/2008JA013884

Woods, T. N., Bailey, S., Solomon, S., Rottman, G., & Worden, J. (1994). Recent solar EUV irradiance measurements. Front Range American Geophysical Union, Boulder, CO

Woods, T. N., Eparvier, F. G., Fontenla, J., Harder, J., Kopp, G., McClintock, W. E., … Snow, M. (2004). Solar irradiance variability during the October 2003 solar storm period. Geophysical Research Letters, 31, L10802. https://doi.org/10.1029/2004GL019571

126

Xu, W., Marshall, R. A., Fang, X., Turunen, E., & Kero, A. (2018). On the effects of bremsstrahlung radiation during energetic electron precipitation. Geophysical Research Letters, 45, 1167–1176. https://doi.org/10.1002/2017GL076510

Yando, K., R. M. Millan, J. C. Green, and D. S. Evans (2011). A Monte Carlo simulation of the NOAA POES Medium Energy Proton and Electron Detector instrument. Journal of Geophysical Research, 116, A10231, doi:10.1029/2011JA016671

Zhang, J.; Dere, K. P.; Howard, R. A.; Kundu, M. R.; White, S. M. (2001). On the Temporal Relationship between Coronal Mass Ejections and Flares. The Astrophysical Journal, 559, 1, 452-462.

Zhang, J., et al. (2007). Solar and interplanetary sources of major geomagnetic storms (Dst ≤ 100 nT) during 1996–2005. Journal of Geophysical Research, 112, A10102. doi:10.1029/2007JA012321

127

Appendix A

List of Acronyms and Abbreviations

ACE-FTS Atmospheric Chemistry Experiment Fourier Transform Spectrometer AE Geomagnetic Index AIMOS Atmospheric Ionization Module Osnabrück Ap Geomagnetic Index BLC Bounce Loss Cone CAM Community Atmosphere Model CCMI Chemistry-Climate Model Initiative CESM Community Earth System Model CIR Co-rotating Interactive Region CMD Central Meridian Distance CME Coronal Mass Ejection CMIP6 Coupled Climate Model Intercomparison Project v6 CO Carbon Monoxide

CO2 Carbon Dioxide CPA Charged Particle Analyzer e* Secondary electron EEP Energetic Electron Precipitation EMIC Electromagnetic ion cyclotron EPP Energetic Particle Precipitation ES Elevated Stratopause EUV Extreme Ultraviolet EVE Extreme Ultraviolet Variability Experiment

128

FISM Flare Irradiance Spectral Model GCR Galactic Cosmic Rays GOES Geostationary Operational Environmental Satellite GPS Global Positioning System H Atomic Hydrogen

H2O Water Molecule + H3O Hydronium ion HALOE Halogen Occultation Experiment

HNO3 Nitric Acid

HOx Odd Hydrogen ISE Impulsive Solar Events Kp Geomagnetic Index LOS Line of Sight MEE Medium Energy Electrons MEPED Medium Energy Proton/Electron Detector MERRA Modern-Era Retrospective Analysis for Research and Applications MIPAS Michelson Interferometer for Passive Atmospheric Sounding MLT Magnetic Local Time MOZART Model for Ozone and Related chemical Tracers MP15 Modified Peck Electron data set N Atomic Nitrogen

N2 Molecular Nitrogen NASA National Aeronautics and Space Administration NCAR National Center for Atmospheric Research NH Northern Hemisphere

129

NO Nitric Oxide

NO2 Nitrogen Dioxide

NOx Odd Nitrogen

NOy Odd Nitrogen and Reservoir Species O Atomic Oxygen

O2 Molecular Oxygen

O3 Ozone OH Hydroxyl radical O1D O-singlet D POAM Polar Ozone and Aerosol Measurement POES Polar-orbiting Operational Environmental Satellite REP Relativistic Electron Precipitation SAGE Stratospheric Aerosol and Gas Experiment SD-WACCM Specified Dynamics version of WACCM SDO Solar Dynamics Observatory SEE Solar EUV Experiment SEM-2 Space Environment Monitor version 2 SH Southern Hemisphere SOLSTICE Solar Stellar Irradiance Comparison Experiment SORCE Solar Radiation and Climate Experiment SPE Solar Proton Event SSW Sudden Stratospheric Warming TEC Total Electron Content TECU Total Electron Content Units TIE-GCM Thermosphere-Ionosphere-Electrodynamic General Circulation Model TIME-GCM Thermosphere-Ionosphere-Mesosphere-Electrodynamic

130

General Circulation Model WACCM Whole Atmosphere Community Climate Model WACCM-D Whole Atmosphere Community Climate Model with D- region ion chemistry UV Ultra-violet XPS XUV Photometer System XRS X-ray Sensor

131