The Test MIST Special Meeting, Southampton, March 26-28th, 2018


MIST Logo_small_0
In its own words MIST is “the community of Magnetosphere, Ionosphere and Solar-Terrestrial researchers working in the United Kingdom. We represent the interests of MIST scientists and hold meetings to showcase MIST science twice a year”.

It is a group which focuses on Space Science, both theoretically and empirically, covering everything from Solar physics [see Shannon’s work] to Planetary Atmospheres, incorporating data from many space-missions, ground-based measurements (for Earth) and models which underpin our current understanding.

MIST holds two meetings a year: Autumn MIST, a one-day meeting; and Spring MIST, a longer 2.5 day meeting (unfortunately these names lead to a lot of pictures of dewy mornings when using Google…).

Each Spring MIST meeting is given a name based on a local geographical feature, this year we were at Southampton University near the River Test. The “Special” is to honour MIST’s 50th anniversary. Less obvious to me was why people kept laughing when the name was mentioned. The only clue was “Any similarity to low frequency emissions on 198 kHz is purely coincidental”. If you like cricket, this may be obvious… but for the rest of us, it’s a reference to the cricket ‘Test Match Special’ radio broadcast. Good thing we’re scientists and not comedians.

This year, it’s safe to say that the Reading delegation took over the meeting. With 9 of us attending (6 presenting!) the Reading Space group was definitely very well represented. Mike Lockwood gave an excellent speech at the conference dinner on the importance of MIST to the community and how he has seen it evolve over the years. See below to read about what everyone presented.

The bi-annual meetings are excellent for keeping up with the current state of the UK’s space-science research as well as maintaining a more informal atmosphere due to the small nature of the community (there were around 60 people at Spring MIST ’18). I find that this all comes together to form a very inviting platform for those of us just starting out in research.

This year’s spring MIST group photo

Shannon Jones presented a poster, “Solar Stormwatch: Using citizen science to investigate CME distortions”.

Coronal mass ejections (CMEs) are the main drivers of hazardous space weather. We are using a novel dataset, created with the help of many citizen scientists through the Solar Stormwatch project, to investigate the effect of the solar wind on these storms. Participants track the shape of CMEs in images from the heliospheric imagers on board the twin STEREO spacecraft, providing an unprecedented level of detail (Barnard et al., 2017). We intend to use this data to extend the work of Savani et al. (2010), looking at how CMEs are distorted under varying solar wind conditions. 

Sarah Bentley gave a talk, “A solar wind-parameterised, probabilistic model of ground-measured ULF waves in Earth’s magnetosphere”.

Large scale ultra-low frequency (1-15 mHz) plasma waves in the magnetosphere are involved in the energisation and transport of radiation belt electrons, a hazardous environment for the satellites underpinning our everyday life. We can construct a statistical model predicting when and where we see these waves in the magnetosphere solely using causally correlated solar wind properties [Bentley et al., 2018]. Unlike existing models, this can be used probabilistically, so that instead of outputting a single value for the power in these waves at each location we can use a probability distribution. Sampling from these distributions turns out to be the best way of predicting total power over a longer event while using the mean values is the best way of predicting the power in the oncoming hour. Using these predicted power values we will eventually be able to predict the effect of these waves on radiation belt particles more precisely over a larger range of the magnetosphere.

download (1)

Oliver Allanson gave a talk, “Particle-in-cell models of diffusion due to whistler mode waves: comparing quasi-monochromatic to broadband waves”.

The momentum space diffusion of electrons due to whistler mode waves is a cornerstone of our current theoretical framework of acceleration (and loss) in Earth’s outer radiation belt. The quasilinear theory of wave-particle interactions provides us with a tractable method to estimate the amount of momentum space diffusion that occurs for a range of wave and ambient plasma conditions. Underlying quasilinear theory is the assumption that waves are broadband, incoherent, and of small amplitude. The right-handed whistler mode manifests in different ways throughout the outer belt: structured chorus, incoherent hiss, near monochromatic transmitter waves, lightning generated whistlers, and large amplitude nonlinear wave packets. It is possible that incoherent hiss is the only example that satisfies all of the formal requirements of quasilinear theory. We use particle-in-cell simulations (the EPOCH code) to model different cases, i.e. from near monochromatic to unstructured broadband, and from small amplitude to large. Through the use of various diagnostics, we explore whether the quasilinear diffusion description is a reasonable description of each case.

Clare Watt gave a talk, “The origin of the whistler-mode spectral “gap” at half electron gyro-frequency in the magnetosphere”.

Near-Earth space contains high-energy electrons, high-energy protons, and a host of different electromagnetic waves that exist over a wide-range of frequencies. Because of the Earth’s magnetic field, and the presence of the high-energy charged particles, the electromagnetic waves do not behave exactly like light in a vacuum, rather they are guided along and across the magnetic field, and interact with the electrons and protons to transport energy and momentum throughout the system. One type of waves are known as whistler-mode waves. They have frequencies of roughly 100-1000Hz and interact with the high-energy electrons in the outer radiation belt. These interactions are thought to be responsible for the energisation of the outer radiation belt. But the waves themselves have many fascinating and mysterious features. Decades of in-situ observations of the waves reveal a persistent frequency gap. Many theories have been put forward to explain the gap, but most rely upon special circumstances that are not guaranteed throughout space. Our recent physics-based simulations reveal a ubiquitous process that can explain the frequency gap, and what’s more, we have identified an independent observational test for the process. Our simulations revealed this new process because advances in computing and simulations allow us to use higher resolution than before – previous work had missed the important fine details of the interaction. At MIST, we reported not only on our simulation results, but also on the recently-published evidence from NASA Van Allen Probes and NASA Magnetosphere Multi-Scale that confirms our independent observational test.

Mike Lockwood gave a talk, “A homogenous aa index”.

Originally complied for 1868-1968 by Mayaud, and extended to the present day by ISGI (International Service for Geomagnetic Indices), the aa geomagnetic index has been a vital resource for studying space climate change over the past 150 years. However, there have been debates about the intercalibration of data from the different measuring stations. In addition, the effect of drift in geomagnetic latitude of the stations, caused by the secular change in the Earth’s field, has not been allowed for. As a result, the components of the aa index for the southern and northern hemispheres have drifted apart. We have corrected for these effects and also for the time-of-day and time-of-year sensitivity of the stations. The resulting indices for the northern and southern hemisphere now agree very closely and the aa index for all years shows a time-of-day and time-of-year “equinoctial” response pattern, as seen in the am index which has been compiled by ISGI from a much larger network of stations since 1959.

Chris Scott gave a talk, “The ionospheric response to intense bombing during World War II”.

There is an increasing number of case studies that demonstrate that the ionosphere can be perturbed from below. The explosion of the chemical plant at Flixborough in 1974 was sufficiently energetic that its effects were detected in the ionosphere (Jones and Spracklen, 1974), lightning has been shown to enhance ionospheric sporadic-E layer electron concentrations (Davis and Johnson, 2005) and there is much interest in the impact of earthquakes on the ionosphere (e.g. Astafyeva et al, 2013). The influence of the troposphere was also cited as the source of unknown variability in modelling work by Rishbeth and Muller-Wodarg (2006). Throughout the second-world war, routine measurements of the Earth’s ionosphere were made at Slough, UK. In this study we will use these data to investigate the impact on theionosphere of various bombing campaigns in order to determine the threshold above which such explosions can be detected in the upper atmosphere.

Presenting in Ponte Vedra, Florida – 33rd Conference on Hurricanes and Tropical Meteorology


You’ve watched many speak before you. You’ve practised your presentation repeatedly. You’ve spent hours, days, months, and sometimes years, understanding your scientific work. Yet, no matter the audience’s size or specialism, the nerves always creep in before a presentation. It’s especially no different at your first international conference!


Between the 16th and 20th April 2018, me, Jonathan Beverley and Bethan Harris were fortunate enough to attend and present at the American Meteorological Society 33rd Conference on Hurricanes and Tropical Meteorology in Ponte Vedra, Florida. For each of us, our first international conference!

Being a regular user of Instagram through the conference, especially the Instagram Story function, I was regularly asked by my friends back home, “what actually happens at a scientific conference”? Very simple really – scientists from around the world, from different departments, universities, and countries, come to share their work, in the hope of progressing the scientific field, to learn from one another, and network with future collaborators. For myself, it was an opportunity to present recently submitted work and to discuss with fellow researchers on the important questions that should be asked during the rest of my PhD. One outcome of my talk for example, was a two-hour discussion with a graduate student from Caltech, which not only improved my own work, but also helped me understand other research in global circulation.

Recordings of the presentations given by University of Reading PhD students can be found at:

Alongside presenting my own work, I had the opportunity to listen and learn from other scientific researchers. The conference had oral and poster presentations from a variety of tropical meteorology subject areas including hurricanes, global circulation, sub-seasonal forecasting, monsoons and Madden-Julian Oscillation. One of the things that I most enjoy at conferences is to hear from leading academics give an overview of certain topic or issue. For example, Kerry Emanuel spoke on the inferences that can be made from simple models of tropical convection. Through applying four key principles of tropical meteorology including the weak temperature gradient approximation and conservation of free-tropospheric moist static energy, we can understand tropical meteorology processes including the Intertropical Convergence Zone, Walker circulation and observed temperature and humidity profiles.

Of course, if you’re going to fly to the other side of the pond, you must take advantage of being in the USA. We saw a SPACEX rocket launch, (just at a distance of 150 miles away,) experienced travelling through a squall line, visited the launch sites of NASA’s first space programs, and explored the sunny streets of Miami. It was a great privilege to have the opportunity to present and attend the AMS 33rd Conference on Hurricanes and Tropical Meteorology, and I am hugely thankful to NERC SCENARIO DTP and the Department of Meteorology for funding my work and travel.


Accurate spectral measurements of solar radiation – why we need improved understanding in the near-infrared

Weather and climate processes are fundamentally driven by energy flows within the Earth-atmosphere system. Incoming solar radiation is absorbed and scattered by gases and aerosols within the atmosphere and absorbed and re-emitted by the Earth’s surface. We therefore need to know how much energy is absorbed by the atmosphere and the height at which this radiation is absorbed.

Currently, we know to reasonable accuracy and precision where most of this energy is accounted for (what we call the global energy budget).

Fig. 1 (Stephens et al. 2012, Nature Geoscience), values in W m-2

Some of the values on the above figure (Figure 1) are highlighted in purple – this indicates that the relative uncertainty (i.e. the range in which this value might plausibly be) on these values is rather high. Reducing the uncertainty on these values is important: this will improve the accuracy of models we use to determine weather and climate. This is achieved by advances in modelling techniques, or in the case of my PhD improvements in available measurements of processes in the atmosphere.

My PhD work focuses on the components circled above, the short-wave atmospheric absorption (i.e. solar energy which is absorbed by the atmosphere as it travels from the Sun toward the surface), and on the incoming solar radiation. The latter of these has a small uncertainty, but this does not quite tell the whole story. The spectral distribution (i.e. at what wavelengths this radiation is emitted) of this energy is also extremely important, since the atmosphere is more transparent at some wavelengths than others.

Fig. 2: Model run of water vapour absorption in the near-infrared for a mid-latitude Summer atmosphere. Everywhere (roughly) above the blue line is considered completely opaque, with varying degrees of transparency below this.

My work focuses on the “near-infrared” spectral region, between about 1-5 μm (or 2000-10000 cm-1) . This region, as can be seen in the above figure, has a “band-window” structure, where parts of the spectrum are completely opaque to radiation, but other parts are almost entirely transparent. Solar radiation with the same wavelength as these band regions where the absorption is strongest will therefore be deposited in the upper atmosphere, while radiation within the windows will be absorbed throughout the atmosphere and by the surface. This structure is almost entirely due to absorption by water vapour.

It is therefore extremely important to characterise the absorption in these windows as much as possible, since any additional absorption will affect where in the atmosphere solar energy is absorbed (unlike additional absorption in the band regions which will barely affect where in the atmosphere this absorption takes place).

Figure 2 also shows the water vapour continuum; a component of absorption which is not currently fully understood. This absorption is a phenomenon not fully accounted for by the theory of water vapour absorption; currently we model it using the MT_CKD model (named such after its creators). The strength of this absorption may be significantly stronger than this model however; laboratory measurements show differences of up to a factor of 100 in the strength at about 1.6 μm!

It is believed (e.g. Radel et al. (2013)) that an increased continuum could contribute about 3 W m-2 to the overall shortwave atmospheric absorption; a significant portion of the 10 W m-2 uncertainty in Figure 1.

My work attempts to resolve this using direct measurements of solar radiation in this 2000-10000 cm-1 region using a Fourier Transform spectrometer, made by the National Physical Laboratory at a site at Camborne, Cornwall, UK. There are a number of challenges making such measurements in the atmosphere; the instrument needs to be properly calibrated, and the conditions in the atmosphere (specifically temperature, pressure, humidity and aerosols) need to be well characterised. This is done using contemporaneous measurements using a radiosonde (to measure the atmospheric profile) and a sunphotometer (to measure aerosol optical depth). These radiosonde measurements are then put into a line-by-line radiation code to calculate the atmospheric optical depth, and these two contributions are subtracted from the total optical depth to get the continuum optical depth.

To derive the continuum it is necessary to know what the incoming solar radiation is. It turns out this is also a significant uncertainty in the literature in the 2000-10000 cm-1 region. While the total incoming solar irradiance is well-known, the distribution of that energy with the electromagnetic spectrum is not so well known. In the spectral region I’m looking at, that uncertainty is about ~10% between different sets of observations.

Since we have direct measurements of the Sun with absolute calibration, we can determine this from our own measurements, and found that the irradiance in this region may be significantly lower (16 W m-2 integrated over the whole spectral region) than expected, which must be made up by contributions elsewhere in the spectrum to account for the small uncertainty in the incoming solar radiation from Figure 1.

Figure 3: Solar spectral irradiance in the 4000-10000 cm-1 region from CAVIAR2 (my work) compared with different sets of observations from other groups (ATLAS3 and Solar2). My work agreed significantly better with the Solar2 work in the 4000-7000 cm-1 region, with good agreement with both in the 7000-10000 cm-1 region. (From Elsey et al. [2017])
Following this, more work was put in to deriving the continuum. This is a more difficult task than simply measuring the incoming solar radiation, since we need to know the different components of the absorption in detail rather than filtering out the effect of the entire atmosphere. Figure 4 shows our best estimate of the continuum, showing a much stronger absorption than MT_CKD. There are large uncertainties however, due to the difficulty in attributing each component of the absorption. Thus, it cannot be ruled out entirely that MT_CKD is representative of the continuum, merely that is likely to be too weak.

Fig. 4: Derived continuum absorption from my observations vs MT_CKD. Dark blue regions indicate k = 1 (67% confidence interval) uncertainties, cyan indicates k = 2 (95% confidence interval) uncertainties.

In the last part of my PhD, I hope to look at what effect these two results might have on the Earth’s energy budget, and look at how much of this 10 W m-2 uncertainty might be accounted for by them. This ties in well with a new project (ASPIC, Advanced Spectroscopy for improved characterisation of the near-Infrared water vapour Continuum) starting up at Reading and the Rutherford Appleton Laboratory in June, which hopes to look at new laboratory measurements of the continuum and assess the effect a strengthened continuum may have on radiation models.


European Geosciences Union General Assembly 2018


Thanks to Hannah Bloomfield, Ambrogio Volonte, and Matthew Lang for photos.

The EGU General Assembly took place from 8th to 13th April 2018 in Vienna and a large group of PhD students and staff from Reading attended. EGU is a large conference covering most areas within geoscience and over 10,000 people attended in all, making it rather overwhelming for a first timer!

The venue: Austria Center Vienna

Each day consisted of five 1.5 hour sessions, interspersed with coffee breaks and lunch. The final session every day was devoted to posters, providing a great opportunity to discuss work and network.

There were many sessions on offer, it was sometimes hard to choose between them. There were oral presentations, posters, and PICOs. I found the PICO sessions an interesting concept. They started with ‘2 Minute Madness’ when all the presenters introduced their topics before the audience were invited to visit the interactive screens to discuss with the presenters. Particularly interesting sessions for me were those on internal gravity waves and forecasting the weather.

Several of the PhD students who made the trip from Reading had talks during the week and the rest had posters. My poster was included in the final session on Friday when everyone was winding down. I was part of the mountain meteorology session and my poster covered the effects of vertical wind shear on gravity wave drag in the Antarctic region. I found it to be a valuable experience as I was able to discuss my work with experts in my field.

Of course it would be a shame to travel all the way to Vienna and not see the sights! As well as having some shorter breaks during the week, I was lucky enough to have the Saturday to explore in the glorious weather before flying back in the evening.

Which solar wind properties drive large-scale plasma waves in Earth’s magnetosphere?

Earth’s radiation belts are a hazardous environment to satellites, which are at risk from the charged particles trapped in near-Earth space. The behaviour of these particles is strongly determined by a spectrum of plasma waves. Ultra-low frequency (“ULF”) plasma waves are large-scale waves with periods on the order of minutes (frequency 1-15 mHz). While these are a fascinating component of near-Earth space, they’re particularly of interest to radiation belt modelling because of their role in the energisation and transportation of radiation belt electrons, so we want to know when and where to expect these waves.

Figure 1: Earth’s radiation belts contain particles (mostly electrons and protons) that are trapped by the Earth’s magnetic field. These need to be understood in order to protect satellites, many of which orbit in the heart of this environment. Missions such as the Van Allen probes (depicted here) provide a way to measure the particle population and the plasma waves which allow for particle acceleration, loss and transport.  Credit: JHU/APL, NASA

These plasma waves are predominantly driven by perturbations of the magnetopause – the boundary between the solar wind and the area dominated by Earth’s magnetic field. A simple example would be a constant tapping on the magnetopause by solar wind pulses – each tap causes a small compression and a magnetic field oscillation (they’re coupled together) which can propagate into the magnetosphere. (Figure 2)

Figure 2: Perturbations at the magnetopause can drive waves that propagate inwards. As the wave travels through the magnetosphere, these oscillations disturb Earth’s magnetic field. We can use the fact that these magnetic disturbances travel along dipole magnetic field lines to measure ULF waves at Earth’s surface.

But can we predict when and where these waves are likely to occur? Since the solar wind is the main driver of ULF waves, we want to be able to predict their effect on electrons from observations of the oncoming solar wind, while most existing models are based on the global geomagnetic activity index, Kp. There are many reasons why this is a poor parameter to base predictions on, the two most relevant being that firstly, it’s a 3-hr averaged index, so we don’t know the value of Kp at the current time (not great for either forecasting or nowcasting) and secondly, it’s so highly derived that it is not really suitable for any kind of statistical description of ULF waves (Murphy et al., 2016).

Previous studies have used a variety of methods to parameterise ULF wave power using solar wind properties (See review in Bentley et al., 2018). It turns out that a difficult part of this question is the solar wind itself. For starters, there is a lot more data describing some conditions than others, e.g. we have far more observations of the solar wind with a speed of 400 km s-1 than 600 km s-1 , and we must account for this if we don’t want our results to be skewed towards the situations where we have more data. But a more difficult problem is the tangled nature of the solar wind properties, which are highly interdependent. (Figure 3) This is partly due to the fact that the solar wind can come from different solar sources, and each one is likely to have a consistent set of properties which then occur at the same time. But also important are the multitude of interactions within the solar wind before it reaches Earth.

Figure 3: Establishing causal relationships is particularly difficult when looking at the solar wind as many properties are highly interdependent. If a quantity D correlates with B and C, is that because they both affect D? Here, only C affects D. But B will still correlate with D because B and C are interdependent. We want to identify only causally correlated parameters.

For example, fast solar wind is generally less dense than the slow solar wind, so speed vsw will anticorrelate with proton number density, Np. But when a region of fast solar wind catches up with some slow solar wind, we will end up with a compression region (Figure 4), so the onset of high speed solar wind will also be related to sudden dense regions and corresponding oscillations of the interplanetary magnetic field (as it folds up due to the compression). If on average we see increased ULF wave power in the magnetosphere when we see high solar wind speeds, is that then due to the speed or due to properties of density or the magnetic field that happen to occur at the same time? Other examples of interdependencies include turbulence, wave interactions and the composition in certain types of solar wind. Many solar wind properties correlate with the speed, because it’s quite a good proxy for all the different types of solar wind.

Figure 4: As fast solar wind catches up with slow solar wind, this creates a compression region ahead and a rarefaction region behind. This is one example of many solar wind interactions that make it difficult to separate the effect of different solar wind properties on the magnetosphere.

Unfortunately most of the existing techniques we might use to construct a parameterisation of ULF wave power on these solar wind properties aren’t appropriate – either they require unphysical assumptions about these interdependencies or they will be difficult to use to investigate the physics behind ULF wave occurrence.

Instead we opted for something simpler – systematically examine all solar wind parameters to find out which ones are causally correlated with ULF wave power. An example of this is shown in Figure 5: take two solar wind parameters to make a grid, and in each bin show the median observed ULF wave power. This allows us to see whether power increases with one parameter when a second is held constant, across different values. This accounts for the interdependence between a pair of parameters and so by systematically comparing many of these plots, we can identify which parameters are causally correlated to power, rather than just correlated to other parameters that affect the wave power. In the example here we can see that when the interplanetary magnetic field Bz component is above zero, ULF wave power increases only with increasing solar wind speed. However, when it’s below zero, ULF power increases with both speed and with more strongly negative Bz.

Figure 5: A two-parameter plot taken from Bentley et al., 2018. We bin the ULF power observed at one station (roughly corresponding to geostationary orbit) at one frequency (2.5mHz) and observe whether it increases with increases in solar wind speed vsw and/or the component Bz of the interplanetary magnetic field. Cut-throughs at constant speed and Bz are shown in (b) and (c).

While this method is very simple, it turns out to be surprisingly powerful – there’s clearly a threshold at Bz=0 that would be averaged over by other techniques, and it also turns out to be the change in proton number density δNp rather than the number density Np that’s causally correlated with power. We can speculate on what physical processes driving the ULF waves are represented by these parameters (see Bentley et al., 2018). It’s likely that the Bz threshold is due to different physical processes that occur when Bz <0, i.e. magnetic reconnection, which I briefly described in a previous blog post.

So by using a simple and systematic method to identify the properties of the solar wind that drive magnetospheric ULF waves, we can resolve three parameters: speed vsw, magnetic field component Bz and proton number density perturbations δNp. Having identified these three parameters opens up new opportunities to model magnetospheric ULF wave power and explore the physics – just when, where and how do we see these waves? And can we quantify how much these parameters contribute – does this change in different regions of the magnetosphere?

Like much of scientific research, answering this one question has opened many more avenues of study to understand these large-scale plasma waves and their role in the dynamics of Earth’s magnetosphere.


Murphy, K. R., I. R. Mann, I. J. Rae, D. G. Sibeck, and C. E. J. Watt (2016), Accurately characterizing the importance of wave‐particle interactions in radiation belt dynamics: The pitfalls of statistical wave representations, J. Geophys. Res. Space Physics, 121, 7895–7899, doi:10.1002/2016JA022618.

Pizzo, V. (1978), A three‐dimensional model of corotating streams in the solar wind, 1. Theoretical foundations, J. Geophys. Res., 83(A12), 5563–5572, doi:10.1029/JA083iA12p05563.

Bentley S.N., C.E.J. Watt, M.J Owens, and I.J. Rae (2018), ULF wave activity in the magnetosphere: resolving solar wind interdependencies to identify driving mechanisms, Journal of Geophysical Research, 123, doi:10.1002/2017JA024740.


The Solar Stormwatch Citizen Science Project

Coronal mass ejections (CMEs), also known as solar storms, are huge clouds of solar material, made up of plasma and magnetic field, emitted from the Sun. If these storms reach the Earth, they can cause geomagnetic storms with severe consequences, such as widespread long-term power-cuts (Canon et al., 2013). Therefore, it’s important to learn as much as we can about the nature and evolution of these storms, to accurately predict if and when they will hit the Earth, and how damaging they will be if they do.

Figure 1: A coronal mass ejection leaving the Sun (Credit: NASA)

For these reasons, the Solar Stormwatch project was created; a citizen science project where volunteers identify and track CMEs in remote-sensing images of space. The original project, jointly created by the Zooniverse and the Royal Observatory Greenwich, asked participants to complete six different activities, and proved very popular. More than 16,000 citizen scientists took part, resulting in seven scientific publications. Based on the success of this, a new version has been created, Solar Stormwatch II, to continue the effort to improve CME forecasts.

Figure 2: A screenshot of the Solar Stormwatch II interface (

In Solar Stormwatch II, volunteers complete an activity called ‘Storm Front’, characterising CMEs in images from the heliospheric imagers (HI) on board NASA’s twin STEREO spacecraft in orbit around the Sun. These imagers take wide-angle images looking from the Sun out into space, and CMEs propagate outwards from the Sun through the field of view. To make the motion of each storm clearer, we show volunteers running difference images, where each image has the previous image subtracted, so only the differences remain. Figure 4 shows example plain and running-difference images for comparison. Each participant is shown three consecutive running-difference images of a CME, and draws around the storm fronts they see in each image, as shown in Figure 5.

Figure 3: An illustration of the twin STEREO spacecraft in orbit around the Sun. (Credit: NASA)
Figure 4a: An example of a running-difference image of a CME. 4b: An example image taken by the heliospheric imagers.
Figure 5: An example set of three consecutive running-difference images of a CME, with storm fronts traced in red.

This is a subjective task, and we don’t expect that everyone will draw storm fronts in exactly the same place. Even two experts might disagree, and these differences could lead to big differences in results (De Koning, 2017). Therefore, we ask 30 people to draw around every storm front, which allows us to combine the observations, find the average location of the storm front, and calculate uncertainties from the distribution of the observations. This makes the dataset more objective and robust than if one expert had created it. Figure 6a shows the average storm fronts and uncertainties found using this method for an example image.

Figure 6a: A differenced HI image with the average storm fronts and uncertainties superimposed. 6b: The average storm fronts for one CME, changing from light to dark red over time, as the CME travels through the field of view.

Typically, researchers only track CME propagation along one slice of each image (Sheeley Jr. et al., 1999); Storm Front allows the whole CME front to be analysed in an unprecedented level of detail (Barnard et al., 2017). This extra detail will allow us to examine how the shape of the CME is distorted as it propagates through the HI field of view (see Figure 6b). Savani et al. (2010) looked at one CME and found that the solar wind, the constant stream of solar material which the Sun emits into space, could explain how the CME shape was distorted; we intend to use the dataset created though Solar Stormwatch II for a statistical comparison between CME distortions and solar wind conditions.

Figure 7: Sign up at

At the time of writing, over 3,000 volunteers have taken part in Solar Stormwatch II, resulting in nearly 60,000 classifications. However, only 30% of the dataset has been completed, so we still need more volunteers. If you’d like to join the effort, please visit and help finish the dataset!

· Barnard et al. 2017 doi:10.1002/2017SW001609
· Canon et al. 2013 doi:1/903496/96/9
· De Koning 2017 doi: 10.3847/1538-4357/aa7a09
· Savani et al. 2010 doi:10.1088/2041-8205/714/1/L128
· Sheeley Jr. et al. 1999 doi:10.1029/1999JA900308

Oceans in Weather and Climate Course 2018


Between the 11th-16th March myself and four other PhDs and post docs attended the Ocean in Weather and Climate (OiWC) course at the Met Office, Exeter. This NERC advanced training course was aimed at PhDs, postdocs and beyond. It provided a great opportunity to spend a week meeting other Oceanography researchers at varying stages of their career, and to expand your understanding of the oceans role in climate beyond the scope of your own work.

The week kicked off with an ice breaker where we had do some ‘Scientific speed dating’, chatting to other participants about: Where are you from? What do you work on? What is your main hobby? What is the biggest question in your field of research? This set the tone for a very interactive week full of interesting discussions between all attendees and speakers alike. Course participants were accommodated at The Globe Inn situated in Topsham, a cute village-sized town full of pastel-coloured houses, cosy pubs, art galleries, and beautiful riverside walks to stretch your legs in the evenings.

The days consisted of four 1.5 hour sessions, split up by caffeine and biscuit breaks to recharge before the next session.

Topics covered in the lecture-style talks included…

  • Dynamical Theory
  • Modelling the Ocean
  • Observations
  • Ocean-atmosphere coupling
  • Air-sea fluxes
  • High Resolution Ocean modelling in coupled forecast systems
  • The Meridional Overturning Circulation
  • The Southern Ocean in climate and climatic change
  • Climate variability on diurnal, seasonal, annual, inter-annual, decadal timescales
  • Climate extremes
  • Climate sensitivity, heat uptake and sea level.
A recurring figure of the week…. taken from Helene Hewitt’s talk on high resolution ocean modelling showing ocean surface currents from HadGEM3-based global coupled models at different resolutions (eddy resolving, eddy permitting and eddy parameterised).


All the talks were very interesting and were followed by some stimulating discussion. Each session provided an overview of each topic and an indication of the current research questions in each area at the moment.

In the post lunch session, there were group practical sessions. These explored observational ARGO float data and model output. The practicals, written in iPython notebooks, were designed to let us play with some data, giving us a series of questions to trigger group discussions to deepen understanding of topics covered that morning.

The course also included some ‘softer’ evening talks, giving research career advice in a more informal manner. Most evenings were spent exploring the lovely riverside walks and restaurants/pubs of Topsham. The final evening was spent all together at the Cosy Club in Exeter, rounding off a very interesting and enjoyable week!