Using the date timing of intense Sporadic E Layer VHF radio reflections to forecast July, summer and Early Autumn weather (temperature) trends by Dr Chris Barnes Bangor Scientific and educational consultants e-mail    May 2015.



 A new method for UK summertime temperature anomaly prediction based on intense VHF sporadic E radio reflection is proposed and validated. Temperature data prediction is available 1-2 months earlier than using other methods. 



The phenomenon of Sporadic E radio reflection has been known about for several decades now, in fact since as far back in history as when  high power VHF TV and FM broadcasting began in the UK and USA and signals were often observed to be received in very anomalous and none line of sight locations [1]. It has also been understood that the phenomenon cannot take place unless there is stratification in the E region of the ionosphere, see Whitehead (1990) [2].      Further, more and more is becoming understood about this propagation mechanism and stratification.    Because Sporadic E is related to multiple triggers involving vertical shear of horizontal travelling ionisation waves in the ionosphere initially of meteoric origin but strongly influenced by planetary tides and waves, gravity waves from the atmosphere below including by jet streams, thunder storms, earthquakes and volcanoes to name a few, I have recently and somewhat remarkably been able to show a complex link between the QBO  ( Quasi –biennial Oscillation) of the equatorial winds and the  spring/ summer start dates of the sporadic E season. The relationship is a complex 4TH order polynomial   and similar complex equations link UK monthly climate anomalies and QBO phase and descent rate, see Barnes 2015 [3].  I have also shown that similar equations hold for an entire season. Figure 1 shows the polynomial equation which predicts the start of the sporadic E season as a function of QBO phase and descent rate together with a compilation function QBO with hindsight summer season temperature anomalies   since 1980.  

Figure 1 : Base polynomial equations showing QBO predicting both retrospective 2m Es propagation and summer season temperature. 

Consider the P Value Results, firstly for the Es predictor: 

We have,   r=.88   DF=10   (10 degrees of freedom) thus the two-tailed P value equals 0.0002.    By  all conventional criteria, this difference is considered to be extremely statistically significant.

Further  and secondly  consider the P Value Results for the summer temperature anomaly  :  P Value Results  r=.563   DF=28  ( 28 degrees of freedom)

            Here, the two-tailed P value equals 0.0012 by conventional criteria, this difference is also considered to be very statistically significant.

  Interestingly the residuals are far greater with negative QBO and there are almost two distinct bands to the data between QBO 0 and -16 which may be solar/NAO/AO effects.  I have discussed these elsewhere [4].  


Effectively, separate start dates can be defined according to MUF (maximum useable frequency).  As far as investigation by radio amateurs is concerned their use of frequency is constrained to the various operating bands which their license permits them to use.   For instance the VHF bands are the 6m band (50-54 MHz), available in all European and American countries, the 4m band (70-70.5 MHz) only available in certain countries and the 2m band ( 144-148 MHz) available in all countries.    Basically the higher the frequency band, the higher needs to be the MUF in order to access the Sporadic E mode of radio reflection/propagation.  Since far more intense ionisation is required at the highest MUF’S these type of conditions only occur about 4-8 times per annum and usually later than the start date as defined for lower MUF’S.


Typically the   6m spring/summer Es   season commences in April but the 2m season not until May, June or even July.



Since both climate anomaly and Sporadic E event dates have been shown to correlate with QBO through sets of polynomial equations there ought also to be some sort of correlation between climate anomaly and sporadic E dates.  Furthermore, the significant features of the two polynomials as plotted above , figure 1, appear to be mirrored across the abscissa or at least approximately mirrored, suggesting that when one function is divided  by the other a  significant degree of linearization ought to take place.   



The start of the Sporadic E season is actually quite hard to define as occurrences at lower MUFS   can be very common. It was decided therefore to define the start of the 2m ( 144 MHz) season instead and also  because using hindsight  data from radio communication journals  it is expected to be far better documented.  Since the Es clouds drift in time and space   it is necessary to define a geographical area for each event. The author is interested in the UK climate, especially Wales.  Generally sizeable sporadic E events affecting the UK will also effect Wales and Eire. 

Only the earliest 2m events effecting these countries at one end of a propagation path were considered in choosing dates to define the first significant event of each year in the study.


The climate anomaly data sets were available at the UK met office website [5, 6].    The first and earliest significant 2m events usually take place in May or June and very occasionally July. It was thus decided to see if the dates of these events could be used in hindsight prediction of July, Summer and Early Autumn temperature anomaly.


The results were plotted using a well-known graph   package, namely Curve Expert by Hyams [7].  The    statistical significance ( p-values)  were obtained using an on –line calculator  from the Regression factors and number of degrees of freedom.   The data set sought after was 1976-2014.  However, since amateur radio data has not been stored on the internet for very long, data regarding 2m Sporadic E events was very sparse in all but the later few of these years, accounting for the differing numbers of degrees of freedom, figure 1.      Furthermore, the UK climate anomaly     is differently recorded by the Met Office  for different years.  In the years 2001 – 2015, graphical and tabulated numeric data is available [5].  However, before 2001 only GIS style mapping is available with a significant    deterioration in accuracy [6]. 


Results and Discussion

Figure 2 shows the result for the Bangor area of Gwynedd and data taken from the GIS maps.  Summer temperature anomaly refers to the average for all three summer months June-August inclusive. 


Figure 2

The regression factor  of .76 with 12 degrees of freedom is highly statiscally significant, with p= 0.0016

Figure 3 shows the method exteneded for the whole of Wales and using the more accurate, 2001- 2014 dataset. 


Figure 3: R=.79


Further   regressions were attempted for the months of July, August and September separately.   A regression factor of .78 was obtained for July, no correlation for August and a factor of .57 (p=.033) for September.

It is thus possible to use the method to predict temperature anomaly for July with some considerable certainty and for the whole of the summer and for the early autumn period but not for August which in North Wales appears to be a most unpredictable month.

Once again ‘banding’ in the residuals is seen which may be inherited from the – ve  QBO data and is possibly  the only thing that limits  this technique.

An independent test of this was to consider if hindsight July temperatures could be used to make similar predictions.   It was    shown that July could predict September temperatures with reasonable certainty but not August.  Yet both strangely and similarly to the case with the Sporadic E method predictor July temperatures are an excellent predictor of the averaged anomaly spread across both August and September, see Figure 4.    


Figure 4, Hindsight prediction of August + September from July anomaly, R=.81


The Sporadic E method is further justified by comparing the monthly regression factors like against like, figure 5. 

Figure 5

Let us consider the P Value Results.    We have r=.99963   DF=3    giving a two-tailed P value is less than 0.0001, considered to be extremely statistically significant. 

In other words the sporadic E 2m start date method has been proved to be a valid climate anomaly prediction method for North Wales Summer Climate temperature anomaly.



Further discussion

It is clear the initial hypothesis is   strongly supported and validated from the point of view of mathematical analysis, but what evidence is there for its physical basis?


Intense sporadic E openings have long been associated with severe thunderstorms.  From personal research I have found jet stream orientation and thunderstorms together are quite critical for the propagation mode.    Davis and Johnson (2005) [8]   have emphasized the notion of Lightning-induced intensification of the ionosphere sporadic E layer. It has been proposed, on the basis of a few observed events that the ionospheric 'sporadic E' layer—transient, localized patches of relatively high electron density in the mid-ionosphere E layer, which significantly affect radio-wave propagation—can be modulated by thunderstorms. They identified a statistically significant intensification and descent in altitude of the mid-latitude sporadic E layer directly above thunderstorms. Because no ionospheric response to low-pressure systems without lightning was detected, they concluded that this localized intensification of the sporadic E layer can be attributed to lightning. They suggested that the co-location of lightning and ionospheric enhancement could be explained by either vertically propagating gravity waves that transfer energy from the site of lightning into the ionosphere, or vertical electrical discharge, or by a combination of these two mechanisms.  

Lightning data, collected using a Boltek Storm Tracker system installed at Chilton, UK, were used to investigate the mean response of the ionospheric sporadic-E layer to lightning strokes in a superposed epoch study. This lightning detector can discriminate between positive and negative lightning strokes and between cloud-to-ground ( CG) and inter-cloud ( IC) lightning. Superposed epoch studies carried out separately using these subsets of lightning strokes as trigger events have revealed that the dominant cause of the observed ionospheric enhancement in the Es layer is negative cloud-to-ground lightning.   In my opinion, this would help account for why not every single thunderstorm causes a sporadic E radio propagation event.

Whitehead (1988) has suggested that  Mid-latitude sporadic-E is most likely due to a vertical shear in the horizontal east-west wind and this theory accounts for the detailed observations of the wind and electron density profiles. Preferred heights of sporadic-E are separated by about 6km and descending layers are often seen moving down with velocities in the range 0.6–4 ms−1. Sometimes sporadic-E layers are very flat and uniform, and at other times form clouds of electrons 2–100km in size moving horizontally at 20–130 ms−1. Sporadic-E is probably not correlated with meteor showers; this is a rather surprising result since the ions are meteor debris.  However, meteoric debris has a very long lifetime in the atmosphere [9].  


Vertically propagating gravity waves from storms can provide Whitehead’s shear.  Sprites have been identified as evidence of vertical gravity wave structures above mesoscale thunderstorms. Large area multicell thunderstorms lead to the formation of vertically oriented cylindrical structures of gravity waves at mesospheric altitudes closely resembling those observed in optical emissions associated with transient luminous glows called sprites.   Taylor (1988) observed a short period gravity wave train was detected by its effect on three upper atmospheric nightglow emissions, the OI 557.7 nm and Na 589.2 nm lines and the OH bands between 715 and 810 nm (Taylor et al., 1987, Planet. Space Sci. 35, 413). Images of these emissions, which were recorded on the evening of 14 August 1980 from the Gornergrat Observatory, Switzerland (45.98°N, 7.78°E), contained high contrast wave-like structures coherent in all three emissions and exhibiting curvature. These properties have been used to identify a thunderstorm centred over southern France as the most likely source of the waves.  Interestingly,   Es clouds are known for their curved surfaces ideal for radio reflection. 


Woodman et al has further elucidated the Es process, finding that wave-like features in range seen on the range/time/intensity (RTI) records of VHF backscatter radars operating in the South of New Zealand are interpreted as being the signature of gravity waves propagating in an ionospheric sporadic-E layer. The data show that, during midsummer in particular, sporadic-E ionisation which has been modified by the passage of a gravity wave can produce two distinct echo types: backscatter from field-aligned irregularities within the sporadic-E layer, probably generated by plasma waves, and a second type of echo resulting from energy backscattered from the surface of the sea after specular reflection in the ionosphere. The backscattering and reflecting region can exist at latitudes at least as low as 49° geographic (57° geomagnetic) latitude during quiet magnetic conditions. They confirmed the patchiness of dense sporadic-E, and concluded that gravity waves at sporadic-E heights have amplitudes of the order of several tenths of a kilometre.  They also concluded that more than likely only Gravity waves with phase fronts parallel to the magnetic dip angle were  capable of producing such distortion in a normally stable and radio inactive E layer , imposing its own temporal and spatial periodicity on the echoes.  This probably additionally accounts for why not all thunderstorms produce sporadic E.  By my own personal experience I have found that thunderstorms tend to be more effective towards the south of an east-west radio propagation path and can sometimes be as much as 200 km south of that path.  Only once have I experienced intense 2m E’s propagation centrally in a UK thunderstorm.   


 Fritts and Nastrom.(1992) [10] considered four cases of mesoscale variance enhancements of horizontal velocity and temperature due to frontal activity, non-frontal convection, and wind shear. These data were obtained aboard commercial aircraft during the Global Atmospheric Sampling Program (GASP) in 1978 and 1979 and from the corresponding meteorological analyses and satellite imagery. Additional GASP data were used to permit a statistical assessment of the importance of various sources of enhanced variances. The results, and those in their companion paper addressing the variance enhancements associated with topography, represent refinements of previous source analyses using the GASP dataset. Significant findings include mean variance enhancements of velocity and temperature due to convection and jet-stream flow ranging from 2 to 8 for 64-km and 256-km data segments, and enhancements for individual segments as high as 20 to 100. The mean 64-km variance enhancement for all variables and source types, relative to a quiescent background, was estimated to be 6.1. These results suggest a major role for localized sources in energizing the mesoscale motion spectrum at horizontal scales < 100 km, and correspondingly greater influences for such motions at greater heights.  KH billows similar to those found in the troposphere are also found in the E-layer.


Others have  shown theoretically that modulation of electron densities in ion layers between 90 and  110 km altitude has been observed using a number of ionospheric diagnostic measurements including scatter of VHF radar waves, artificially pumped optical emissions, scintillations of satellite beacon transmissions. Kelvin–Helmholtz (K–H) turbulence driven by a sheared wind profile is a strong candidate for the source of these modulations. A two-dimensional numerical model is used to calculate the nonlinear evolution of ion layers in ionosphere near 100 and altitude in response to neutral turbulence driven by a wind shear. The amplitude of a K–H billow is allowed to grow as a linear perturbation on the neutral atmosphere to a level that is 10% of the wind shear. The time dependent model of the ionosphere responds to neutral wind perturbation initially by imposing a quasi-sinusoidal modulation near the altitude of the ion layer. This is followed by compression of the initially stratified layer into structures with the wavelength of the K–H instability. These structures are uniform strips in the meridian perpendicular to the direction of the zonal wind. Near, where the ion gyro frequency (ωi) is about equal to the ion collision frequency, the equilibrium solutions are clumps at the altitude of the shear. Near, two stable, rippled layers are produced with a  given separation. The amplitudes of the density modulations in the ion layers vary by as much as 500% throughout the simulation. The simulations illustrate the complex evolution of the ion layer structures from small-amplitude, K–H wind turbulence.


Existing theory of the stability of a stably stratified fluid containing a strong vertical shear suggests that unstable waves may develop when the curvature of the velocity profile changes sign and the Richardson number is somewhere less than 1/4. Some observations are described which show the properties of atmospheric billow clouds formed in travelling amplifying waves (transverse to the shear vector), on occasions when these conditions appear to be met. Static instability seems to arise in parts of the wave-pattern where layers are inverted, and to cause a convective overturning, which may halt the wave development. The most pronounced waves occur in the upper troposphere in association with jet streams, in layers of strong wind shear which are usually dry. They probably only rarely produce clouds, and may more frequently be responsible for the clear-air turbulence encountered by aircraft. The associated relative air velocities occur over a range of scales: up to about 1 km in the convective regions, and up to the several km associated with the billow wave-lengths, with magnitudes of up to 10 m sec−1 or more.  Of course jet streams are another E’s trigger long since acknowledged by radio hams. Clearly their upwardly propagating billow initiated gravity waves are essential as thunderstorms.


Such billows are found in reality by analyzing the field-aligned coherent radar backscatter observed, for example, this was done over Gadanki, India (13.5°N, 79.2°E), with a narrow beam pointing almost vertically, Choudhary et al (2005) [11] present convincing experimental evidence for the presence of low-latitudes tilted sporadic ionization layers close to 10 km in vertical extent that move horizontally through the field of view of the radar. Using the data from high temporal (3 s) resolution experiments, we also show that the line-of-sight Doppler velocities associated with at least some of the quasi-periodic striations have very clear vortex-like structures cutting across a vertical plane inside regions of strong horizontal wind shears. The power as well as the Doppler width peak together, and they often reach their peak values near the centre of a vortex, where the magnitude of the Doppler velocity is minimum. The Doppler properties and spatial distribution of the 3 m echoes are explained in terms of a local electro dynamical process that makes ions and electrons move with the vertical neutral wind. Both the wind field and the tilt of the layers are in turn consistent with the presence of Kelvin-Helmholtz billows. Billows themselves are triggered by a shear instability in the large ambient zonal wind; strong zonal wind shears clearly have to be present when sporadic E layers are observed. In our case, the breaking of an originally uniform and horizontal sporadic E layer into tilted pieces aligned more or less parallel to one another, and their motion through the radar field of view in the presence of a mean zonal wind, give the echoes their quasi-periodic appearance.  Here the link between QBO and E’s is re-affirmed.


Let us re-visit lighting. In a warmer climate more lightning is to be expected. Price and Rind (1994) and Price (2008) [12] have made the analysis.  They use the Goddard Institute for Space Studies (GISS) general circulation model (GCM) to study the possible implications of past and future climate change on global lightning frequencies. Two climate change experiments were conducted: one for a 2×CO2 climate (representing a 4.2°C global warming) and one for a 2% decrease in the solar constant (representing a 5.9°C global cooling). The results suggest a 30% increase in global lightning activity for the warmer climate and a 24% decrease in global lightning activity for the colder climate. This implies an approximate 5–6% change in global lightning frequencies for every 1°C global warming/cooling. Both intra-cloud and cloud-to-ground frequencies are modelled, with cloud-to-ground lightning frequencies showing larger sensitivity to climate change than intra-cloud frequencies. The magnitude of the modelled lightning changes depends on season, location, and even time of day.  The notion of increased cloud to ground lightening is particularly relevant to the case in hand, we have already seen above it is negative cloud to ground lightening which has been associated with sporadic E enhancement.  

It is instructive to consider the behaviour of lightning in Europe, see Figure 6 below. 

eFigure 6

The peak month for commencement of the 2m Es season is June.  The average date for commencement is June 3rd.   Thus we may regard this date as being related to 10% of the duration of this month can be regarded as having a 10% increase in flashes over May assuming linearity in the cumulative frequency slope. The highest positive anomaly within the period was 2.6 C and the lowest negative anomaly was -0.5 C.    Based on Price and Rind’s calculation, 3.2 C represents a 19.2 day advancement of the start to the season i.e. circa May 14th.   A .5 C negative anomaly represents a 2 day retardation. In the real data set for North Wales, the largest advancement in 2m Es season start date noted was some 26 days and the largest retardation was some 18 days.    Thus the hypothesis that summer climate temperature prediction by following E’s start date as a result of earlier temperature   enhanced lightening activity is strongly supported.


Possibly some other factor could also be at work. Rishbeth   (1990) [13] has considered the possibility of a greenhouse effect in the ionosphere.   Following a suggestion by Roble and Dickinson that increases in the mixing ratios of mesospheric carbon dioxide and methane will cool the thermosphere by about 50K, their paper examines the consequences for the ionosphere. The cooling and the associated composition changes, as described by Roble and Dickinson, would lower the E- and F2-layer peaks by about 2 km and 20 km respectively, but changes in the E- and F2-layer electron density are small.  It is uncertain if such small height changes in the E layer prior to stratification would be sufficient to changes E’s start date significantly.

In any event at least for the moment, anthropogenic changes do not seem to have invalidated the method. 


Further work

It is hoped to investigate the relationship between Sporadic E and rainfall in another publication in due course.



A new method for UK summertime temperature anomaly prediction based on VHF sporadic E radio reflection  has been proposed and validated. Although marginally less accurate than using July temperature data prediction is available 1-2 months earlier. 





3.      RadCom Plus, Vol. 1, No. 1 pp 22-30 May 2015