This article reviews atmospheric changes associated with 44 solar eclipses, beginning with the first quantitative results available, from 1834 (earlier qualitative accounts also exist). Eclipse meteorology attracted relatively few publications until the total solar eclipse of 16 February 1980, with the 11 August 1999 eclipse producing the most papers. Eclipses passing over populated areas such as Europe, China and India now regularly attract scientific attention, whereas atmospheric measurements of eclipses at remote locations remain rare. Many measurements and models have been used to exploit the uniquely predictable solar forcing provided by an eclipse. In this paper, we compile the available publications and review a subset of them chosen on the basis of importance and novelty. Beyond the obvious reduction in incoming solar radiation, atmospheric cooling from eclipses can induce dynamical changes. Observations and meteorological modelling provide evidence for the generation of a local eclipse circulation that may be the origin of the ‘eclipse wind’. Gravity waves set up by the eclipse can, in principle, be detected as atmospheric pressure fluctuations, though theoretical predictions are limited, and many of the data are inconclusive. Eclipse events providing important early insights into the ionization of the upper atmosphere are also briefly reviewed.
This article is part of the themed issue ‘Atmospheric effects of solar eclipses stimulated by the 2015 UK eclipse’.
The weather during solar eclipses has always been of interest, because of the effect it has on the viewing experience. Early observations were compiled by Halley  to deduce the path of a total solar eclipse across the UK on 22 April 1713, with the help of observers across the country. Inclement weather occasionally impeded the spectacle, as Halley explained, ‘My worthy colleague Dr John Keill by reason of Clouds saw nothing distinctly at Oxford but the End…’. As well as the excitement of darkness falling, and birdsong stopping, Halley noted other widely observed meteorological effects, particularly ‘the Chill and Damp which attended the Darkness of this Eclipse’.
The first quantitative eclipse weather observations of which we are aware, on 30 November 1834, were reported in the Boston Medical and Surgical Journal . Although not named, the author was a diligent observer, finding that temperature measured in the shade and, ‘of course, in a northern exposure’, dropped by 1.5°F from the start of the eclipse to greatest obscuration, with no substantive change in the pressure. Soon afterwards, careful qualitative observations from a UK partial eclipse of 15 May 1836 were made by Birt , who used Luke Howard’s then-novel cloud classifications to give a detailed account of cloud changes during the eclipse. Cumulus (‘fair-weather’) clouds died away as the eclipse developed, which might now be explained in terms of diminished convection from reduced solar heating.
Systematic eclipse weather campaigns were pioneered by Winslow Upton, who was professor of astronomy at Brown University . He led a series of expeditions, with his first eclipse meteorology publication from the Russian solar eclipse of 19 August 1887  followed by other eclipses over the USA [5,6]. His work is characterized by multi-instrumented sites with several observers focusing on different aspects to maximize the scientific return (automated records were not used). Soon after this, Clayton  provided a modernizing influence, being the first to apply meteorological techniques such as use of ‘anomalies’, i.e. subtracting the average behaviour to identify the changes caused by the eclipse. Clayton was also the first to recognize the special circumstances provided by an eclipse in imposing a well-understood diminution of solar radiation. As Clayton put it, ‘the eclipse may be compared to an experiment by Nature … by eliminating the influence of other known phenomena’, which he used to good effect by investigating the dynamical effects of eclipses, discussed in more detail in §2a.
By the end of the twentieth century, improvements in computing and atmospheric physics led to the extensive use of meteorological models. The first application of a meteorological model to an eclipse was published by Gross & Hense  in anticipation of the 11 August 1999 European total solar eclipse. Since then, there have been several sophisticated model studies of eclipse meteorology, which will be covered in §2b.
The final effect of eclipses on the weather to attract significant scientific attention is the generation of gravity waves, fluctuations in atmospheric pressure generated by the atmospheric cooling from the obscuration of the Sun by the Moon. This is a relatively recent concept, first predicted by Chimonas  and Chimonas & Hines . Anderson & Keefer  planned and carried out pressure measurements on the basis of these predictions, and retrospectively analysed older eclipse pressure data for the possible existence of gravity waves. Attempts to measure gravity waves at subsequent eclipses have been relatively common, but most of the findings have been inconclusive and inconsistent (§2c).
(a) Summary of previous work and structure of the paper
Atmospheric measurements during 44 eclipses are summarized in table 1. Many articles exist describing eclipse meteorology, about 120 in total (figure 1). Publications have become significantly more numerous since the 16 February 1980 eclipse, with the 11 August 1999 European total solar eclipse provoking the most reports. The number of papers written about each eclipse in recent years appears to be related to the eclipse’s path—if it is over densely populated areas such as Europe, North America, India and China, the Moon’s shadow will pass over existing meteorological stations or provoke atmospheric field campaigns. The role of the amateur observer can also be significant, and is discussed by Hanna .
It is not possible to summarize all the results in table 1 here (a subset of the data are tabulated by Kameda et al. ), so the approach taken is to focus on work that is of particular interest, usually the first of its kind or a novel method. For example, meteorological measurements of eclipses over sparsely populated areas such as the poles remain of scientific interest owing to the low Sun angles and opportunities to investigate rarely observed effects [77,87], and these are discussed in §2d. This paper is organized by topic up to, but not including, results from the March 2015 eclipse, covering circulation changes (§2a), meteorological modelling (§2b) and gravity waves (§2c) in the troposphere and stratosphere. Other effects, such as atmospheric electricity, and novel approaches, such as measurements aloft, are mentioned in §2d. In the upper atmosphere, here defined as 100 km upwards, photoionization is dominant and therefore eclipses produce changes that are more significant than those seen in the troposphere and stratosphere. An entire paper could be written on historical work investigating the response of the upper atmosphere to eclipses, but we have provided a brief summary for completeness (§3).
2. Meteorological effects of eclipses in the troposphere and stratosphere
The obvious effect of an eclipse is that the solar radiation reaching the Earth decreases as the Moon obscures the Sun’s disc. In a total solar eclipse, the solar radiation drops to zero; in a partial eclipse, it does not reach zero. (There is negligible difference between the atmospheric effects of partial and annular eclipses of similar magnitude, as an annular eclipse is when the Moon covers only the centre part of the Sun, leaving a ring of it showing, whereas during a partial eclipse the Moon leaves a crescent of the Sun showing.) Cooling from solar eclipses is greater when the Sun is higher in the sky, because the Sun’s obscuration has a larger relative effect on the downwelling solar radiation, i.e. at local noon and near the midsummer solstice. For example, a substantial temperature change, −5°C, was recorded after the 21 June 2001 total solar eclipse over northern Zimbabwe (15°S), which took place near local noon, but close to the southern winter solstice (figure 2). The maximum surface temperature changes in the literature are about −7°C . For the eclipse in figure 2, the lag between totality and the temperature minimum is 30 min, similar to Vogel et al.  but longer than most other clear sky lags (e.g. Kameda et al.  quoted 6–30 min with an average of 12 min). This lag is thought to be caused by the thermal inertia of the surface layer  but it has also been reported to be empirically related to the global solar radiation Sg at fourth contact . The sensitivity of the temperature change to solar radiation is 0.017±0.001°C/(Wm−2), consistent with other clear sky data .
Figure 3 is a summary of meteorological effects of partial solar eclipses over Reading University’s Atmospheric Observatory, which we believe to be the most comprehensive set of eclipse weather observations from a single site. The diminution in solar radiation is usually detectable in the global solar radiation measurements even under thick cloud, as for the two March eclipses shown, but for the 12 October 1996 eclipse, no solar radiation change is evident. Both the August eclipses, which were similar in terms of time of year, day and cloud, showed a clear reduction in surface temperature. The largest temperature drop was 1.5°C, on 11 August 1999, owing to the 97% magnitude of the eclipse and the broken cloud cover. In comparison, the cloudy umbral region at Camborne, UK, had a local temperature minimum before eclipse maximum .
Near-surface relative humidity changes are typically anticorrelated with near-surface air temperature changes, and increase directly as a consequence of the cooling during eclipses. For example, Namboodiri et al.  reported a maximum variation of 19% with a 28 min time lag after the maximum eclipse associated with a 3.5°C cooling during the 2010 annular eclipse over Thumba, India. A reduction of specific humidity has also been reported and attributed to eclipse-induced subsidence of drier air , but relative humidity changes are typically dominated by the temperature effect.
Atmospheric cooling by eclipses is responsible for most of the other more subtle meteorological changes discussed below. Circulations can be modified, small pressure waves, ‘gravity waves’ set up, and atmospheric turbulence suppressed. These can affect surface winds, temperatures and radiation, and the electrical state of the atmosphere.
(a) Circulation changes
Eclipses often modify synoptic or mesoscale circulations in favour of their own local cooling-induced effects. Associated with these circulation changes may be an ‘eclipse wind’ which has been reported anecdotally as darkness descends, perhaps related to the heightened emotional state of some eclipse observers . Subrahmanyam et al.  found that the sea breeze circulation at Thumba, a coastal site in southern India, was suppressed during a relatively long 77% annular eclipse occurring near local noon on 10 January 2010. During maximum eclipse, radiosonde measurements indicate that the sea breeze cell was only about 300 m thick compared with 600 m under similar synoptic conditions the preceding day.
Clayton  was the first to propose that solar eclipses modified atmospheric circulation, based on a compilation of observations from several sites across the USA during the 28 May 1900 solar eclipse, as was mentioned in §1. His results suggested a cooler anticyclonic (clockwise) circulation extending about 1500–2000 miles from the umbra, with a cyclonic (anticlockwise) circulation ring beyond, about 1000 miles across, surrounded by an outer ring of higher pressure. Modelling studies by Prenosil  predicted similar effects to those of Clayton’s theory for the 1999 eclipse over Europe. In high-resolution wind measurements in the umbra and penumbra of the 11 August 1999 eclipse, Aplin & Harrison  found that the eclipse circulation was roughly consistent with Clayton’s model for the penumbra, but that only the umbral region acted as a cold-cored cyclone, with a cold outflow into a surrounding anticyclonic region. Subsequent studies comparing observations from the Met Office surface station network with a forecast from its unified model (which did not represent the eclipse) found that the circulation changes seen were likely to be eclipse-induced . These circulation changes could contribute to the sensation of an ‘eclipse wind’ reported by some observers.
(b) Modelling studies
There are a range of model studies of eclipses, summarized in table 2. The two main categories are discussed here: first, meteorological models (often from national weather services), which represent most of the recent output, covered in §2b(i), and second, a few solar radiation models covered in §2b(ii).
(i) Meteorological models
One of the underlying problems of eclipse meteorology is distinguishing changes caused by eclipses from other, independent, atmospheric effects. This situation is most complicated when meteorological changes are occurring simultaneously with an eclipse, for example when a weather front is arriving at the observation site. The standard approach used is (i) to compare the data with days before and after the eclipse or (ii) to interpolate the changes during the eclipse period to estimate what would have happened if there was no eclipse. As pointed out by Gray & Harrison , neither of these approaches is ideal, because (i) requires consistent conditions on the days either side of the eclipse and (ii) neglects temporal lags in eclipse-induced changes, which may persist after the eclipse. Under these circumstances, in particular, meteorological models can play an important role, as was hinted in §2a.
Gross & Hense  used the (now no longer operational) German Weather Service forecast model with 14 km grid spacing to predict the effect that the 1999 solar eclipse trajectory would have had over Central Europe for the weather conditions on 11 August 1998, a clear and hot day, and found that a near-midday eclipse six weeks away from the summer solstice would cause a substantial (3.5°C on average) and rapid temperature drop, and that the atmospheric circulation would be affected. Unfortunately, as Aplin & Harrison  reported, a weather front in North Western Europe meant that the weather, and observed effects, could differ substantially from clear sky values. Prenosil  used a low-resolution (63.5 km grid spacing) regional weather prediction model to simulate the tropospheric response to the total solar eclipse on 11 August 1999 over central Europe both with real-time data for that day and for a water vapour and cloud-free scenario. A cold-cored cyclone circulation was expected, and the maximum temperature change of −2°C was predicted in an area north of Dover, UK, and east of Calais, France; both predictions were not inconsistent with the observations by Aplin & Harrison . Vogel et al.  also performed numerical experiments for the 11 August 1999 eclipse, using a high-resolution (4 km grid spacing), non-hydrostatic regional atmospheric model (KAMM, Adrian & Fiedler ) coupled to a surface vegetation model and positioned over the upper Rhine valley. An idealized, spatially and temporally constant, background flow and cloud-free conditions were assumed. Temperature decreases of 6–7°C were simulated. The modification of winds on slopes in the inhomogeneous terrain led to large changes in wind direction. However, for homogeneous terrain, small effects on wind field were found, with ‘no implication of the so-called eclipse wind’. Founda et al.  performed numerical model experiments with the weather research and forecast model for the 29 March 2006 eclipse. They used nested domains with 2 km grid spacing in the highest-resolution domain. In this domain, they predicted eclipse-induced 1.5 m temperature anomalies that were consistent in magnitude and timing with those observed (e.g. predicted cooling of 2.8°C in the centre of Athens near eclipse maximum compared with 2.6°C observed). Some wind speed and direction changes were measured in totality, but Founda et al.  attributed them to a combination of the synoptic evolution and changes in the local sea breeze flow.
Eckermann et al.  used a US Department of Defense model with a ‘high top’ of 0.001 hPa for the eclipse of 4 December 2002. This model, including a radiative code to predict the absorption of atmospheric species such as water vapour and ozone, was used to predict surface temperature and pressure changes by differencing runs initialized for identical pre-eclipse conditions with and without the eclipse trajectory included. Surface temperature changes were greatest over land, with a cooling of 4°C, consistent with observations for other eclipses. Surface pressure fluctuations of 0.1–1 hPa were also predicted from gravity waves (similar to the 0.3 hPa predicted by Prenosil ), which will be discussed in the next section. A further application of meteorological modelling demonstrated for this work is that atmospheric effects owing to eclipses can be predicted in remote areas, such as for the path of the 2002 eclipse over southern Africa, the southern Indian Ocean and Australia. To the best of our knowledge, this eclipse yielded no meteorological observations owing to its trajectory over sea and sparsely populated areas.
One of the problems of modelling eclipses is that a numerical weather model run with and without an eclipse gives results that represent the eclipse in the model, but do not verify the model itself. To move away from this, Gray & Harrison  employed a different methodology. A high spatial and temporal resolution model was initialized with pre-eclipse weather conditions and run with no knowledge of the eclipse. This model run, when compared with observations made at a network of UK Met Office stations, and those in Aplin & Harrison , could be used to verify the model’s prediction of the changes caused by the eclipse over the UK. Over a subselected penumbral region excluding the southwestern peninsula (chosen to correspond to the clearer sky, and better instrumented region), a reduction in wind speed of 0.7 ms−1 and a backing of the wind direction, i.e. an anticlockwise change corresponding to a more easterly direction, were attributed to the eclipse. This wind direction change, while clearly not associated with an actual gust of wind, also supports the concept of an ‘eclipse wind’, although it is not clear which, if any, of these effects would be detectable by a human observer.
(ii) Solar radiation models
In addition to standard meteorological models, eclipses provide an ideal opportunity for testing solar radiation models. These models mainly simulate the path of photons in the eclipsed region. While this situation is relatively well understood for partial eclipses , total eclipses present technical challenges, because photons detected within the umbra must have arrived by multiple scattering, and, therefore, three-dimensional scattering code is needed. These models have contributed to understanding of mainly small effects such as the effect of aerosols on the visible radiation and the contribution of the solar corona to the radiation . Limb darkening, changes in the spectral irradiance across the Sun’s disc, can also be predicted during solar eclipses using radiative models .
(c) Gravity waves
Gravity waves are atmospheric pressure waves, first predicted as a consequence of the transient cooling induced by a solar eclipse by Chimonas  and Chimonas & Hines . Specific predictions were made for the 7 March 1970 eclipse, and, in response, Anderson et al.  made sensitive pressure measurements and surveyed historical data. Many of the historical pressure measurements reviewed showed fluctuations of about 24 Pa (1 Pa=0.01 hPa) lasting for about half the duration of the eclipse. The measurements made by Anderson et al.  showed a periodicity of 89 min with magnitude of 25 Pa, consistent with the earlier work. These fluctuations were an order of magnitude greater than the 0.2 Pa changes predicted by Chimonas . More recent predictions are larger, 10–100 Pa [75,104], and the range of periodicities actually detected during solar eclipses, summarized in table 3, are from 20 min up to several hours.
The variability in both measurements and theoretical predictions originates from the range of atmospheric layers from which eclipse-induced gravity waves can be generated, and because several mechanisms are involved. In the stratosphere and troposphere, solar UV radiation is converted into heat (infrared radiation) by ozone and water vapour, respectively. When the solar UV radiation is attenuated, at night or during a solar eclipse, these layers of the atmosphere are cooled, and a gravity wave is induced. The daily gravity wave is known as the atmospheric solar tide; further gravity waves are therefore expected to emanate from the penumbral and umbral regions of solar eclipses . Several theories exist explaining exactly how these gravity waves are generated and the effects that can be expected from them, but limited predictions are available for surface pressure changes ( and references therein), and the existing observations are only approximately consistent both with each other and with the theoretical work. Measurements made so far are summarized graphically in figure 4.
(d) Other measurements
In this section, other eclipse weather measurements of interest from table 1 are briefly reviewed.
(i) Atmospheric electricity
The first measurements to be made at eclipses beyond the standard meteorological quantities of pressure, temperature, wind, humidity and solar radiation were of atmospheric electricity. Atmospheric electricity refers to both thunder and lightning, and the subtler electrical properties of the atmosphere away from thunderstorms, such as the air’s electrical conductivity. In the early years of the twentieth century, atmospheric electricity was a scientific priority owing to the recent discovery of radioactivity, and its ability to ionize the air. It was well known at the time that natural ionization led to the slight electrical conductivity of air, but the origin of ‘penetrating radiation’ which ionized air through thick layers of lead would not be fully understood until Hess’s discovery of cosmic rays in 1912 . Although the full text of Knoche & Laub’s  report on atmospheric electrical measurements during eclipses is not available, solar effects on atmospheric electricity would have clearly been of scientific interest and may have motivated early eclipse experiments. An additional source of atmospheric electrical observations was from magnetic observatories, such as the magnetic data from the 21 August 1914 total solar eclipse compiled by Bauer & Fisk , which includes atmospheric electrical and meteorological measurements. Bennett  gives a more detailed account of atmospheric electrical measurements during solar eclipses.
(ii) Measurements aloft
The first successful measurements away from the surface during a solar eclipse were made by Samuels , who arranged for six kites, each carrying a meteograph, to fly from several sites in the penumbra of a solar eclipse over the USA on 24 January 1925. As the eclipse occurred at approximately 0800 local time, when the Sun was low in the winter sky, only two of the six flights showed any detectable response. The largest temperature change was detected in North Dakota, where the eclipse magnitude was 95%, at a height of 400–800 m. (The earliest attempt at eclipse measurements aloft was the launch of two balloons before and after maximum obscuration from Eskdalemuir Magnetic Observatory in the UK on 21 August 1914, but the balloons were lost, so no data were retrieved .) No further above-ground measurements during eclipses were attempted, as far as we know, until instrumented rocket launches in the 1970s . In the twenty-first century, radiosonde (meteorological weather balloon) launches during eclipses have become more popular [67,97,114] with some limited recent rocket launches .
(iii) Measurements near the poles
As discussed in §2, measurements of the atmospheric effects of eclipses are most common over populated areas. The unpopulated environments of the poles receive little attention; for example, Kameda et al.  provide evidence that 18 total solar eclipses over Antarctica were completely unobserved since the continent was first visited by Captain Cook in the 1770s. In what were probably the first polar meteorological measurements during a total solar eclipse—and almost certainly the first human sighting of any Antarctic eclipse—Kameda et al.  and co-workers from the Japanese Antarctic Research Expedition observed the eclipse of 23 November 2003 from Dome Fuji (77°S, 3810 m). The temperature of −51°C dropped by 3°C in response to the eclipse, although the maximum solar radiation in cloudless skies was only 200 Wm−2. No clear effects on atmospheric pressure or wind were detected. While the temperature change observed was typical, the time lag between the end of totality and the temperature minimum was longer for this eclipse than for others, which motivated Kameda et al.  to compare the time lag with minimum temperature against the clear sky global solar radiation for several eclipses, as in §2.
The eclipse measurements closest to the poles were published by Sjöblom , who used five sites near Longyearbyen on the Svalbard archipelago at 78°N. The eclipse of 1 August 2008 was of particular interest, because, although partial (93%) at that site, the midnight Sun was obscured, so the eclipse had a significant effect because of the lack of diurnal variation in solar radiation. During and after the eclipse, katabatic winds started to flow down to the valleys linking to the large fjord in which Longyearbyen is located, after which stratus clouds started to form, which advected onto land as fog. This fog persisted for 3 days, and all air traffic into and out of the archipelago was stopped. There were few explanations for the fog other than the eclipse, because there were no synoptic changes, nor, of course, a sunrise or sunset to modify the local surface radiation balance. Downslope winds are known to occur in mountainous environments as a consequence of local cooling similar to that induced by an eclipse . This eclipse has therefore also been the eclipse with the greatest economic impact, even though relatively few people were in the path of the Moon’s shadow.
3. Eclipse effects on the upper atmosphere
There have been extensive studies of the Earth’s ionosphere and upper atmosphere during solar eclipses, as it was recognized early in the twentieth century that an eclipse could provide a useful test of scientific theories explaining the behaviour of the upper atmosphere, here defined as from about 100 km altitude upwards.
(a) Scientific motivation
The ionosphere is a weakly charged plasma layer of the upper atmosphere, largely generated when atmospheric gases are ionized by incoming solar extreme ultraviolet and X-ray radiation and lost through recombination reactions. At any given time, the rate of change of the ion number concentration, N, in a given volume can be represented by the expression 3.1where q is the ion production rate, L is the ion loss rate and T is the loss rate through ion transport.
We now know that, in the ionospheric E region (at around 100 km), the dominant neutral gases are molecular. When ionized, these molecular ions recombine with free electrons, and any excess electron energy causes the neutral molecule formed to break up in a process known as dissociative recombination. The average lifetime of an ion in the E region is of the order of seconds and so the transport term in equation (3.1) is effectively zero. At the higher altitudes of the F2 region (about 300 km), the dominant ion species is O+, which, owing to the lower gas densities and the atomic ions’ inability to recombine with energetic electrons, has a longer average lifetime (approx. 10 s) and transport becomes important. The intervening F1 layer therefore represents a transition region between molecular and atomic chemistry.
Before spaceflight or even sounding rockets could reach such lofty altitudes, the exact composition of the upper atmosphere remained unknown, and solar eclipses were seen as an opportunity to measure the ionospheric decay rate in order to determine which species were involved.
Techniques were first developed to determine the ion number density using radio . A radio signal is returned from the ionosphere when it reaches an altitude where the local plasma frequency matches the radiofrequency. Because the local plasma frequency is proportional to the square root of the ion number density, transmitting radio pulses with a range of shortwave frequencies revealed the height profile of free electrons, with the time of flight of the signal being used to estimate the height of the ionospheric layers.
During a solar eclipse, with ion production removed at totality, the loss rate, L, was estimated from the time delay between totality and the subsequent minimum in atmospheric ionization. These experiments invariably underestimated the loss rate due to the fact that, while emissions from the visible solar disc were obscured by the Moon at totality, emissions from the solar corona were not. Using eclipse experiments to determine loss rates fell out of fashion after sounding rockets were developed that could make direct measurements of upper atmospheric composition. Now that these loss rates are well known, these early eclipse experiments have been re-interpreted to estimate how the relative size of the solar corona has varied throughout the twentieth century [116,117]. These two papers contain references to the original research publications, along with details of previously unpublished eclipse measurements obtained from international data archives.
Early studies were not restricted to investigations of the loss rate. Beynon & Brown  list the details of 22 eclipse experiments prior to 1956 and report the details from a symposium held at the Royal Society in London, UK, in August 1956 in which experiments conducted during eclipses in the 1950s were discussed. Topics included studies of solar radiation, solar radio noise, ionospheric recombination and eclipse-induced geomagnetic effects.
Modern eclipse studies have benefited from the advent of advanced radar facilities [119,120] and have concentrated on radio propagation studies , wave propagation  and the impact of solar eclipses on the upper atmosphere as a way of verifying complex atmospheric models [123–125].
Solar eclipses offer a unique opportunity to measure the atmospheric effects of a well-defined attenuation of the incoming solar radiation, an aspect first noted by Clayton , who combined multiple observations to make pioneering deductions about circulation changes from eclipses. Clayton’s theory was not universally accepted at the time [13,126] and while more recent work has found eclipse-induced circulations consistent with Clayton’s theory [26,105], other work has found either no circulation changes  or attributed these changes to changes in the synoptic situation or in local mesoscale flows such as sea breezes or mountain slope flows [82,98,99,102]. Clayton’s work also represented a transition from the purely empirical and observational aspects of pre-1900 eclipse weather to the twentieth-century approach of synthesizing measurements and searching for broader atmospheric effects beyond surface changes in solar radiation and temperature. In the first half of the twentieth century, solar eclipses offered a much-needed opportunity to study the upper atmosphere, which was then almost totally inaccessible, and in the second half, the focus shifted to understanding the gravity waves produced by eclipses and their propagation through the atmosphere. The use of models has now become widespread in eclipse meteorology, both to predict the atmospheric changes caused by eclipses and to present a verification opportunity for the model.
Eclipse meteorology is growing as a subject with each eclipse over populated areas now attracting many publications, most notably the 1999 and 2006 eclipses over Europe. There are now over 100 scientific papers reporting highly diverse aspects of eclipse meteorology and it has not been possible to cover them all here, beyond summarizing pioneering work and the significant atmospheric effects. The major open scientific question in eclipse meteorology is the origin and effect of gravity waves, which are still poorly understood with many conflicting theories and observations. The detailed circulation changes from eclipses, particularly their extent (or not) under cloudy skies, also require further investigation. It is likely that the forthcoming 2017 total solar eclipse passing from west to east over the USA will stimulate further work in this area.
Data are available on request from the corresponding author.
K.L.A. carried out the data analysis and literature search, and prepared the plots and tables. All authors contributed to manuscript drafting and review, and gave the final approval for publication.
We declare we have no competing interests.
We received no funding for this study.
Dr J. Milford (formerly of Department of Physics, Zimbabwe University) provided the data shown in figure 2. Prof. R.G. Harrison (Department of Meteorology, University of Reading) provided the data shown in figure 3. Eclipse ephemera were obtained from http://www.timeanddate.com/, http://calsky.com/ and http://eclipse.gsfc.nasa.gov/eclipse.html.
One contribution of 16 to a theme issue ‘Atmospheric effects of solar eclipses stimulated by the 2015 UK eclipse’.
- Accepted November 18, 2015.
- © 2016 The Author(s)
Published by the Royal Society. All rights reserved.