Vol52,3,2009 301 ANNALS OF GEOPHYSICS, VOL. 52, N. 3/4, June/August 2009 Key words Ionospheric propagation – mid-latitude trough – radiolocation – propagation modelling – backscatter sounding 1. Propagation along the mid-latitude trough 1.1. Introduction For terrestrial HF radio systems, the elec- tron density depletion in the trough region re- duces the maximum frequency that can be re- Aspects of HF radio propagation E. Michael Warrington (1), Alain Bourdillon (2), Eulalia Benito (3), Cesidio Bianchi (4), Jean-Philippe Monilié (3), Martha Muriuki (1), Marco Pietrella (4), Veronique Rannou (3), Hanna Rothkaehl (5), Stephane Saillant (3), Ozgur Sari (6), Alan J. Stocker (1), Ersin Tulunay (7), Yurnadur Tulunay (8) and Nikolay Y. Zaalov (1) (1) Department of Engineering, University of Leicester, UK (2) IETR, Université de Rennes 1, France (3) Electromagnetism and Radar Department, Onera, Palaiseau, France (4) Istituto Nazionale di Geofisica e Vulcanologia, Roma, Italy (5) Space Research Centre PAS 00-716 Warsaw, Poland (6) Aselsan A.S., Macunkoy, Ankara, Turkey (7) ODTÜ/METU, Department of Electrical and Electronics Engineering, Ankara, Turkey (8) ODT /METU, Department of Aerospace Engineering, Ankara, Turkey Abstract The propagation characteristics of radio signals are important parameters to consider when designing and operating radio systems. From the point of view Working Group 2 of the COST 296 Action, interest lies with effects associ- ated with propagation via the ionosphere of signals within the HF band. Several aspects are covered in this paper: a) The directions of arrival and times of flight of signals received over a path oriented along the trough have been examined and several types of propagation effects identified. Of particular note, combining the HF obser- vations with satellite measurements has identified the presence of irregularities within the floor of the trough that result in propagation displaced from the great circle direction. An understanding of the propagation effects that result in deviations of the signal path from the great circle direction are of particular relevance to the operation of HF radiolocation systems. b) Inclusion of the results from the above mentioned measurements into a propagation model of the norther- ly ionosphere (i.e. those regions of the ionosphere located poleward of, and including, the mid-latitude trough) and the use of this model to predict the coverage expected from transmitters where the signals impinge on the northerly ionosphere. c) Development of inversion techniques enabling backscatter ionograms obtained by an HF radar to be used to estimate the ionospheric electron density profile. This development facilitates the operation of over the hori- zon HF radars by enhancing the frequency management aspects of the systems. d) Various propagation prediction techniques have been tested against measurements made over the trough path mentioned above, and also over a long-range path between Cyprus and the UK. e) The effect of changes in the levels of ionospheric disturbances on the operational availability at various data throughput rates has been examined for the trough path mentioned earlier. The topics covered in this paper are necessarily brief, and the reader is referred to full papers referenced herein on individual aspects. Mailing address: Prof. E. Michael Warrington, Depart- ment of Engineering, University of Leicester, Leicester, LE1 7RH, U.K.; e-mail: emw@leicester.ac.uk Vol52,3,2009 20-09-2009 19:06 Pagina 301 302 E.M. Warrington, A. Bourdillon, E. Benito, C. Bianchi, J.-P. Monilie, M. Muriuki, M. Pietrella, V. Rannou, H. Rothkaehl, S. Saillant, O. Sari, A. J. Stocker, E. Tulunay, Y. Tulunay and N. Y. Zaalov flected by the ionosphere along the great circle path (GCP). For long paths, the signal is often received via a ground/sea-scatter mechanism to the side of the GCP (Stocker et al., 2003). For shorter paths, gradients in electron density as- sociated with the trough walls and embedded ionospheric irregularities often result in propa- gation in which the signal path is well dis- placed from the great circle direction, with di- rections of arrival at the receiver offset by up to 100° (Rogers et al., 1997). Deviations from the great circle direction impact not only on radi- olocation systems for which estimates of a transmitter location are obtained by triangula- tion from a number of receiving sites, but also on any radio communications system in which directional antennas are employed. Further- more, the Doppler and multi-mode delay spread characteristics of the signal are also af- fected when propagation is via scatter/reflec- tion from irregularities in or close to the north wall of the trough (Warrington and Stocker, 2003). Recently, measurements have been made of the direction of arrival (DOA) and amplitude as a function of time of flight (TOF) of HF signals received over two paths oriented along the trough between Uppsala, Sweden (2001 and from August 2006 to January 2008) and Helsin- ki, Finland (since December 2006) and the Uni- versity of Leicester’s field site near to Leicester (see fig. 1). The first of these periods was close to sunspot maximum, and the second close to sunspot minimum. 1.2. Characteristic types of propagation along the trough Examination of the observations along the trough path from Uppsala and Helsinki has en- abled five types of off great circle propagation 18−Apr−2007, 22UT, Kp=0 20 oW 10 oW 0 o 10 oE 20 o E 30 o E 48 oN 54 oN 60 oN 66 oN 72 oN Leicester Uppsala Helsinki No rth wa ll Sou th w all 18−Apr−2007, 22UT, Kp=6 20 oW 10 oW 0 o 10 oE 20 o E 30 o E 48 oN 54 oN 60 oN 66 oN 72 oN Leicester Uppsala Helsinki North wall Sou th w all Fig. 1. Maps showing the paths between Uppsala and Leicester and Helsinki and Leicester together with a typ- ical position for the trough, as estimated by the model of Halcrow and Nisbet (1977) for 0000 UT on 11 March 2006 for Kp values of 0 and 6. The four lines indicate the outer and inner edges of the north and south walls of the trough. Vol52,3,2009 20-09-2009 19:06 Pagina 302 12 13 14 15 16 17 18 19 20 21 22 23 00 0 10 20 30 40 50 60 ex11, Upp, 10 MHz, 2007/04/18−2007/04/19, Type II A zi m u th ( d e g re e s) 12 13 14 15 16 17 18 19 20 21 22 23 00 4 5 6 7 8 T im e o f fli g h t (m s) Time (UT) 303 Aspects of HF radio propagation events to be identified (see Stocker et al. (2009) for full details): • Type 1: These are characterised at onset by a large sudden increase in time of flight accom- panied by a sudden deviation of the bearing to the north. Subsequently, the time of flight de- creases over a period of hours while the az- imuth either remains fairly constant or returns slowly to close to the great circle direction. Type 1 events were commonly observed, ex- cept in the summer, close to solar maximum (2001) (see Siddle et al., 2004a for a number of examples) when they also tended to be longer- lived and were observed on a wider range of frequencies. Theoretical considerations indicate that these events are consistent with the signal scattering from irregularities in the poleward wall of the trough or from the auroral region (Siddle et al., 2004b). • Type 2: The time of flight gradually increas- es and is accompanied by a gradual deviation of the azimuth to the north (e.g. see the traces for 10 MHz in fig. 2). The rate of change of time of flight and azimuth tends to be fairly constant un- til, after a period of up to a few hours, the signal is no longer detected. • Type 3: The time of flight gradually in- creases and is accompanied by either a gradual southward (Type 3S) or, less frequently, north- ward (Type 3N) deviation of the azimuth. After a few hours either the signal is no longer detect- ed or the azimuth slowly returns to the great cir- Fig. 2. Type 2 example, azimuth and TOF only. Vol52,3,2009 20-09-2009 19:06 Pagina 303 304 E.M. Warrington, A. Bourdillon, E. Benito, C. Bianchi, J.-P. Monilie, M. Muriuki, M. Pietrella, V. Rannou, H. Rothkaehl, S. Saillant, O. Sari, A. J. Stocker, E. Tulunay, Y. Tulunay and N. Y. Zaalov cle direction (this distinguishes Type 3N propa- gation from Type 2). • Type 4: This type of propagation is charac- terised by a sudden large increase in the time of flight and a strong deviation to the south (typi- cally the azimuth is around 120°). The signal is usually relatively weak, and is consistent with sidescatter from a ground reflection in the vicinity of the Alps. • Type 5: This type is similar to Type 1, ex- cept that although there is a sudden increase in the time of flight, the signal is deviated to the south instead of the north. Over the following few hours the azimuth usually returns to the great circle direction. The occurrence (as a percentage of nights of observation) of Types 1, 2, and 4 as a func- tion of season, frequency and Kp is presented in fig. 3 (Types 3 and 5 are generally rare). The seasons have been defined with spring includ- ing all of March and April, summer including May to August, autumn September and Octo- ber and winter November to February (as in ITU, 1997). Type 1 events are most common in spring with no clear trend with frequency. For 4.64 MHz, this type of propagation is also 4.64 8.01 11.12 0 10 20 30 40 Type 1 Summer Autumn Winter Spring 4.64 8.01 11.12 0 10 20 30 40 O cc u rr e n ce ( % o f n ig h ts ) Type 2 4.64 8.01 11.12 0 10 20 30 40 Frequency (MHz) Type 4 4.64 8.01 11.12 0 10 20 30 40 Type 1 0−2 2−4 4+ 4.64 8.01 11.12 0 10 20 30 40 O cc u rr e n ce ( % o f n ig h ts ) Type 2 4.64 8.01 11.12 0 10 20 30 40 Frequency (MHz) Type 4 Fig. 3. The percentage occurrence of Types 1, 2, and 4 off great circle propagation (August 2006-September 2007). Left hand panels show variation with season, and the right hand panels the variation with Kp (averaged 1800-0000 UT). Vol52,3,2009 20-09-2009 19:06 Pagina 304 305 Aspects of HF radio propagation common in winter, while for most other fre- quencies it occurs least often in winter. This behaviour may be contrasted with that ob- served in 2001 (see Siddle et al., 2004a,b) where Type 1 propagation was frequently ob- served (~40–70% of nights depending on fre- quency) except in summer. Furthermore, the events observed in 2001 tended to consist of larger changes in time of flight and azimuth and to be of longer duration. Type 2 events tend to occur more frequently in spring and, for frequencies in the range 8.01-11.12 MHz, in the summer. However, for 6.95 MHz, there is little other seasonal variation, while at 4.64 and 14.36 MHz the events are rarely observed. In spring, the occurrence strongly depends on fre- quency, peaking at 10.39 MHz. Type 4 events are also a springtime phenomenon, being ob- served on over 30% of nights for 10.39 and 11.12 MHz, but rarely at other frequencies. At frequencies of 8.01 MHz and below, Type 1 propagation events become more fre- quent as Kp increases, while for frequencies higher than this the percentage of nights on which this type of propagation is observed is roughly independent of Kp (although there is Fig. 4. Map showing the inferred reflection point of an 8.01 MHz signal on 18 April 2007, 2030UT to 19 April, 0000 UT. Type 2, 8.01 MHz, 2007/04/18, 21:17 UT, Kp=2.5 20 oW 10 oW 0 o 10 oE 20 o E 30 o E 45 oN 55 oN 65 oN Uppsala Leicester Vol52,3,2009 20-09-2009 19:06 Pagina 305 306 E.M. Warrington, A. Bourdillon, E. Benito, C. Bianchi, J.-P. Monilie, M. Muriuki, M. Pietrella, V. Rannou, H. Rothkaehl, S. Saillant, O. Sari, A. J. Stocker, E. Tulunay, Y. Tulunay and N. Y. Zaalov some suggestion of a slight decrease in occur- rence with increasing Kp). For Type 2 propaga- tion, there is a clear decrease in the occurrence of observations with increasing Kp for all fre- quencies except 4.64 MHz, where the opposite is the case. Type 4 propagation is more com- monly observed at the middle range of frequen- cies (8.01-11.12 MHz), and becomes more fre- quent with increasing Kp. 1.3. Comparison of satellite measurements with HF measurements The mechanism by which Type 1 off great circle path propagation occurs has been investi- gated through simulation (Siddle et al., 2004b). However, the modelling has not yet been able to explain why the deviations of dif- ferent types occur on one day but not on other- wise very similar days. In previous work, the parameters describing electron density irregu- larities embedded in the background iono- sphere used in simulation have not been well defined. In this paper, the HF measurements are examined in conjunction with satellite observa- tions of the electron density structure. Some caution must be exercised in comparing the HF and satellite measurements since the satellite al- titude is about 700 km, while the HF propaga- tion is most strongly affected by the ionosphere at heights of about 200-400 km. Since the elec- tron density distribution in the topside iono- sphere is the near-Earth signature of the magne- tospheric plasmapause, it may be expected that the structure of the irregularities due to precipi- tations will be well correlated at different alti- tudes, while the behaviour of the instabilities Fig. 5. The electron density and temperature measured on board of DEMETER satellite on 18 April 2007. Vol52,3,2009 22-09-2009 14:20 Pagina 306 307 Aspects of HF radio propagation inside the trough region will depend on height. For reasons of space, only a single example of Type 2 off great circle propagation from 18 April 2007 will be presented here. Based on the measured angles of arrival and the time of flight of the HF signal (fig. 2), and assuming a single, mirror-like reflection (see Siddle et al., 2004b), the reflection point (fig. 4) moves steadily north of the great circle path reaching a point close to the poleward wall of the trough before the signal is no longer re- ceived. It is evident from Figure 4 that the re- flection points of the HF signal in this event are located inside the trough. The DEMETER satellite data (fig. 5) indicates that, as well as the expected reduction in electron density, there are strong filamentary electron density struc- tures (i.e. irregularities) inside the main ionos- pheric trough at this time, and we suggest that scattering from these irregularities is the main mechanism responsible for formation of Type 2 off great circle propagation. 2. Ionospheric model for ray-tracing Zaalov et al. (2003; 2005) have developed a unique ray-tracing model that accurately re- produces many of the features observed in the experimental measurements referred to earlier to a level well beyond that which we originally anticipated would be possible. A major out- come of these ray-tracing simulations is that paths other than those that have been the subject of experimental investigation can readily be as- sessed. The simulations make use of a numerical ray tracing code (Jones and Stephenson, 1975) to estimate the ray paths through a model iono- sphere comprising two Chapman layers, the main parameters of which (critical frequency, critical height, vertical scale height of each lay- er) are based on values obtained from the Inter- national Reference Ionosphere (IRI) (Bilitza, 1990). The most important causes of off-great circle propagation in the polar cap are the pres- ence of convecting patches or sun-aligned arcs of enhanced ionisation. Patches are formed in the dayside auroral oval and generally convect in an anti-sunward direction across the polar cap into the nightside auroral oval, whereas arcs occur under different geomagnetic conditions and drift in a duskwards direction. Localised, time varying, perturbations in the electron den- sity are then applied to the background model to represent the convecting patches and arcs of enhanced electron density. Sun aligned arcs (Carlson et al., 1984): the shape of each sun-aligned arc is defined within the model by a small number of three-dimen- sional Gaussian perturbations in electron densi- ty of different spatial scales (altitude, longitude and latitude) randomly distributed near to the centre of the arc. Several Gaussian perturba- tions were combined in defining the shape of each modelled arc in order to prevent the shapes of the arcs being too stylised. For all arcs away from close proximity to the dawn or dusk auro- ral oval, the plasma strands are elongated for several hundreds or thousands of kilometres with a latitudinal scale which is significantly larger than the longitudinal scale. Evolution of the structures relative to the propagation path is determined by the rotation of the Earth beneath the arcs and by movement of the arcs in the dawn-dusk direction. Convecting patches (Weber et al., 1984; Buchau et al., 1983): the temporal evolution of the patches relative to the propagation path is simulated by means of a convection flow scheme coupled with the rotation of the Earth beneath the convection pattern, the precise form of which depends upon the components of the IMF. The intensity and spatial scales of the patches can also be varied with simulation time. In practice, the shape, size and number of patches in the convection flow area depends up- on many geophysical parameters, not only upon the instantaneous values but also upon their his- tory. By using up to four vortices based on the modelled convection flow patterns associated with the various IMF orientations many realis- tic situations may be simulated. Mid-latitude trough: the Halcrow and Nis- bet model (1977) was used in the simulations as a basis for the position of the trough walls. In order to add smaller-scale structure to the wall, the following modifications were performed: (a) The latitude of the walls was perturbed by two-scaled random functions of longitude, and Vol52,3,2009 20-09-2009 19:06 Pagina 307 308 E.M. Warrington, A. Bourdillon, E. Benito, C. Bianchi, J.-P. Monilie, M. Muriuki, M. Pietrella, V. Rannou, H. Rothkaehl, S. Saillant, O. Sari, A. J. Stocker, E. Tulunay, Y. Tulunay and N. Y. Zaalov (b) a landscape of patches elongated in the di- rection of the trough were added to each wall. Initially, no perturbations were added to the floor of the trough. The depletion of the trough maximum was set according to Kp, typically, 30% for Kp = 2-3 and 60% for Kp = 6. Particle precipitation: the auroral oval is an enhancement of electron density caused by par- ticle precipitation in the E-region and above. As a function of distance along (near-vertical) field lines, the density enhancement was modelled as starting 100 km from the ground, having one or more peaks of about 1013 electrons/m3 around 110 km, and then decaying slowly toward 200 km (Bates and Hunsucker, 1974). Small scale electron density enhancements can also be added to the model. The model is able to reflect the day-to-day variation of received HF signals. See, for exam- ple, the time history of the azimuth of arrival and time of flight of a 10.4 MHz signal propa- gating along Uppsala to Leicester path on 18-19 April 2007 given in fig. 2 and the simulation re- sults in fig. 6. Fig. 6. Simulated time history of the azimuth of arrival and time of flight of a 10.4 MHz signal propagating along Uppsala to Leicester path between noon 18 April and noon 19 April 2007. Azimuth deviation - top pan- el, time of flight - bottom panel. Vol52,3,2009 20-09-2009 19:06 Pagina 308 309 Aspects of HF radio propagation 3. Area coverage simulations The area coverage to be expected from a transmitter at a given location can be estimated by ray-tracing through the model ionospheres described in Section 2. A large number of rays launched in an azimuth/elevation grid from the transmitter are traced through the model iono- sphere, and the signal strength at the receiver estimated by determining the ray-density in the area around the receive antenna. For the exam- ple simulations presented in this case, the dy- namic range shown in the figures has been re- stricted to 20 dB in order to highlight the modal structure of the signal. An example outcome of this process is illus- trated in figs. 7 and 8. The first of these figures shows a modelled electron density distribution at a height of 200 km and a time of 0230 UT (the dayside ionosphere is positioned in this figure around the bottom left quadrant). Patches of en- hanced ionisation have been randomly posi- tioned within the model ionosphere, as described previously. The outcome of the ray-tracing process is il- lustrated in fig. 8 for a frequency of 10.4 MHz for transmitters located at Iqaluit and Tromsø. The upper frames indicate the expected area coverage when the presence of the patches is ignored, and the lower frames for when the Fig. 7. Example electron density distribution over the high latitude region. The dayside ionosphere is around the lower left hand quadrant. Patches of enhanced electron density are also included. Vol52,3,2009 20-09-2009 19:06 Pagina 309 310 E.M. Warrington, A. Bourdillon, E. Benito, C. Bianchi, J.-P. Monilie, M. Muriuki, M. Pietrella, V. Rannou, H. Rothkaehl, S. Saillant, O. Sari, A. J. Stocker, E. Tulunay, Y. Tulunay and N. Y. Zaalov patches are taken into account. Focussing of the rays at the 1, 2 and 3 hop skip zone ranges is clearly evident in all cases. It is interesting to note that the presence of the patches severely distorts the background pattern on the ground, that coverage is reduced in places but also that coverage is obtained in areas where it was not present without the presence of the patches. This later situation is more evident at higher frequencies, not illustrated here. In considering the effect of the presence of the patches, it is important to remember that the patch distribution shown is only illustrative and that in reality the patches will be distributed dif- ferently, and also that as time progresses the patches will move in accordance with the pre- vailing convection cell pattern (in turn, a func- tion of the geomagnetic conditions). 4. Inversion of HF radar backscatter ionograms HF radars use the refraction by the iono- sphere of radio waves emitted by the radar to de- tect targets up to 3000 km. The distance covered by the radio waves depends on different parame- ters: the ionospheric characteristics, the transmit- ted frequency and the elevation angle. In order to optimise the radar frequencies for the geograph- ical area of interest and to localize the targets ac- curately, knowledge of the ionospheric charac- Fig. 8. Calculated signal strength variations for a 10.4 MHz transmitters located at Iqaluit (left frames) and Tromsø (right frames) covering the high latitude region using the modelled ionosphere of fig. 7. The upper frames are without the presence of polar patches, and the lower frames are with the presence of the patches. Vol52,3,2009 20-09-2009 19:06 Pagina 310 311 Aspects of HF radio propagation teristics is required in quasi real time. In to ac- quire the necessary ionospheric information, the radar can be used as an oblique backscatter sounder. For this purpose, an inversion method is needed as the forward problem is strongly non linear. By scanning the radar beam in elevation for a fixed azimuth at one operating frequency, the group path can be measured to obtain a backscatter ionogram (a 3D image of magnitude, elevation angle β, and group path P’). The Quasi-Parabolic Segment model (QPS) developed by Dyson and Bennett (1988) has been chosen for the inversion. In this model, each layer of the ionosphere and also the join- ing segments can be characterized with a QP segment by only three parameters: the critical frequency (fc), the peak height (hm) and the semithickness (ym). Then the group path P’ and the ground range D can be obtained using ana- lytical propagation equations. 4.1. Inversion technique The purpose is to recover the initial model parameters (fc, hm, ym) of each layer of the iono- sphere from n points of the backscatter ionogram (n�3). With these model parameters the calcula- tion of the ground range, D, is possible so the re- al position of the target can be determined. The data points used are the coordinates of n group paths (P’meas 1, …, P’meas n) at n fixed elevation angles (βref 1, …, βref n) called of reference for each layer. The method, based on Tarantola (1987) and Landeau et al. (1997), will invert each ionosphere’s layer one by one. For each pa- rameter a value is chosen in the parameter space to create a QPS model of the ionosphere and to simulate the elevation-group path curve by using ray tracing. The coordinates of the n group paths simulated corresponding to the n fixed elevation angles are taken (P’simu 1, …, P’simu n). With the coordinates of the n group paths measured and simulated, the a posteriori proba- bility density, σp, can be calculated, using the equation: , , , , f h y P P f h y P 1 1 p c m m meas i simu i c m m meas ii n 1 σ δ = + -= l l l ^ ^ d h h n % where δP’meas1, …, δP’measn represent the vari- ances of the measurement errors over P’meas1, …, P’measn. These variances are presumed to be Cauchy distributed and independent of each other. The optimal parameters are those for which σp is maximum. To find them, each value of the parameters space must be evaluated. But this procedure is costly in time and the ionosphere is changing with time, sometimes on a time scale smaller than 15 minutes. It is thus neces- sary to use an optimization algorithm to get a good approximation of the global optimum. Two different optimization algorithms have been tested: the simulated annealing and the ge- netic algorithm. Simulated annealing: The simulated anneal- ing algorithm (SA) developed by Kirkpatrick et al. (1983) is based on the manner in which met- als recrystallize in the process of annealing. For each step of the SA algorithm a random neigh- bour of the current solution is considered. If the new solution is better, it is chosen. If it is worse, it can still be chosen with a probability that de- pends on the difference between the correspon- ding function values and on a parameter called the temperature (T). The temperature is gradually decreased dur- ing the process. At the beginning the current so- lution changes almost randomly, but the accept- ance of bad solutions decreases as T goes to ze- ro. The allowance for bad solutions saves the method from becoming fixed at local minima. Genetic algorithm: The second optimization method used is a genetic algorithm developed by Goldberg (1989). Genetic algorithms are based on the mechanics of natural selection and genetics. An initial generation is created where the individuals are binary coded strings. By us- ing selection, crossover and mutation a new generation is created from the old generation. Each generation is better than the previous one, until the optimal solution is found. 4.2. Validation on synthesized data Synthesized data are created from a model of electronic density profile given by predic- tions of the ionosphere (table I). After calculat- Vol52,3,2009 20-09-2009 19:06 Pagina 311 312 E.M. Warrington, A. Bourdillon, E. Benito, C. Bianchi, J.-P. Monilie, M. Muriuki, M. Pietrella, V. Rannou, H. Rothkaehl, S. Saillant, O. Sari, A. J. Stocker, E. Tulunay, Y. Tulunay and N. Y. Zaalov ing the group path-elevation angle curve, we add a zero mean Gaussian noise (7.5 km and 2° of standard deviation) to simulate a real backscatter ionogram that is used to test the in- version method. The results of the inversion method are compared to the initial values on fig. 9 and in table I. In fig. 9 the group path-elevation curve obtained by inversion match with the original curve used as data. The two optimization methods are com- pared in fig. 10 and in table I. The genetic algo- rithm converges faster than the simulated an- nealing (1.72 minutes vs 15 minutes with a PC Intel Dual Core). Furthermore, the estimated parameters provided by the genetic algorithm are closer to the initial values than by using the simulated annealing. In future works, the genet- ic algorithm will be used. 4.3. Results on real data The inversion method has also been tested on real data using an elevation-scan backscatter ionogram, recorded on 11 June 2007 at 2000 UT. In this particular sounding, only the F lay- er was present (nighttime’s profile). For valida- tion purposes vertical ionograms have been col- lected by the Ebre ionosonde which was select- ed because it is located in the great circle of the oblique sounding backscatter beaming. Figure 11 compares the elevation angle-group path ionograms obtained with the data inversion and the vertical ionosonde inversion. The corre- sponding electron density profiles presented in fig. 12 show a general agreement of the 2 profiles. 5. Comparison between observations and prediction techniques 5.1. Time of flight measurements and their application to testing predictions methods that approximate to ray tracing Accurate predictions of the main parameters characterising the HF ionospheric channel re- quire a detailed knowledge of the ionospheric conditions that only numerical or analytic ray- tracing techniques can provide. Numerical ray- tracing models (such as Jones and Stephenson, 1975 and Norman et al., 1994) are very accu- rate but they have a high computational time. Analytic ray-tracing techniques that enable ray tracing through horizontal gradients along and in the direction of the ray path (Norman and Cannon, 1997; 1999) are computationally less intensive than numerical ray tracing. For this reason their use is particularly advantageous in HF applications such as in the real-time fre- quency management of OTH (over-the-hori- zon) radar systems (Coleman, 1998). Three different predictions methods that ap- proximate ray-tracing techniques were tested: (a) IRI-95 based on the monthly median elec- Table I. QPS model parameters, initial and estimated, comparison. Layer Parameter Initial value Estimated value Estimated value with SA with GA E fc 2.62 MHz 2.59 MHz 2.58 MHz E hm 105.4 km 109.4 km 109.24 km E ym 15.1 km 19.28 km 19.17 km F fc 7 MHz 6.99 MHz 7 MHz F hm 261.8 km 263.36 km 262.9 km F ym 93.4 km 94.75 km 93.55 km Vol52,3,2009 20-09-2009 19:06 Pagina 312 313 Aspects of HF radio propagation Fig. 9. Backscatter ionogram inversion results. Fig. 10. Comparison of the optimization algorithms speed. Vol52,3,2009 20-09-2009 19:06 Pagina 313 314 E.M. Warrington, A. Bourdillon, E. Benito, C. Bianchi, J.-P. Monilie, M. Muriuki, M. Pietrella, V. Rannou, H. Rothkaehl, S. Saillant, O. Sari, A. J. Stocker, E. Tulunay, Y. Tulunay and N. Y. Zaalov Fig. 11. Elevation angle-group path curves obtained with the inverted parameters and with the ionosonde in- version parameters, and data points used for the inversion at a low operating frequency. Fig. 12. Electron density profiles obtained with the inverted parameters and with the ionosonde parameters. Vol52,3,2009 20-09-2009 19:06 Pagina 314 315 Aspects of HF radio propagation tron density profiles provided by the IRI-95 model (Bilitza, 1990; 2001), (b) SIRM&BR_D based on the SIRM (Zolesi et al., 1993; 1996) in conjunction with the Bradley-Dudeney mod- el (Bradley and Dudeney, 1973), and (c) ICEPAC based on the Ionospheric Communica- tions Enhanced Profile Analysis and Circuit prediction program (Stewart, undated). For the IRI-95 and SIRM&BR_D predic- tion methods, the monthly median electron den- sity profiles provided by the models have been used to calculate the vertical plasma frequency profiles at the mid-point of the Uppsala - Leicester radio link (see Section 1). As the ef- fect of Earth’s curvature is important for ground ranges greater than 500 km, the curvature of the Earth and the known ground range (1411 km) of the radio link have been taken into account for calculating the angle of incidence of the ray at the base of the ionosphere. The values of the angle of incidence have been calculated assum- ing a very simple geometry based on ionospher- ic reflections taking place from a simple hori- zontal mirror at the appropriate height. Subse- quently, the secant law was applied to calculate the oblique transmission frequencies. The length of the oblique ray path, its corresponding time of flight (TOF) and take off angle have been calculated from simple geometry. The version of ICEPAC used in this study utilizes the monthly median electron density profiles (Haydon and Lucas, 1968) at the mid- point of the radio link derived from CCIR foF2 and M(3000)F2 coefficients (CCIR, 1966) and, from these, provides predictions of elevation an- gle, TOF, virtual height, etc. for a given oblique transmission frequency and time of day. In order to test the validity of these methods under quiet ionospheric conditions, compar- isons between the predictions and the TOF me- dian measurements for the one hop propagation modes 1E and 1F were carried out for different frequencies and seasons. For 1E modes (see table II), IRI-95 and SIRM&BR_D provide a similar performance, while ICEPAC is in general somewhat better. This is particularly the case when Es is present. The errors for 1F modes (table III) are larger Table II. The minimum and maximum value of the r.m.s. error of TOF are shown taking into account all the frequencies which arrive at the receiver following the 1E on great circle propagation modes. Winter months Equinoctial months Summer months 1E propagation mode r.m.s.min – r.m.s.max r.m.s.min – r.m.s.max r.m.s.min – r.m.s.max (ms) (ms) (ms) IRI-95 0.038 – 0.099 0.044 – 0.085 0.013 – 0.093 SIRM&BR_D 0.045 – 0.099 0.043 – 0.084 0.008 – 0.072 ICEPAC 0.002 – 0.048 0.022 – 0.062 0.002 – 0.082 Table III. The minimum and maximum value of the r.m.s. error of TOF are shown taking into account all the frequencies which arrive at the receiver following the 1F on great circle propagation modes. Winter months Equinoctial months Summer months 1F propagation mode r.m.s.min – r.m.s.max r.m.s.min – r.m.s.max r.m.s.min – r.m.s.max (ms) (ms) (ms) IRI-95 0.071- 0.325 0.072 – 0.393 0.111 – 0.572 SIRM&BR_D 0.097– 0.139 0.070 – 0.254 0.108 – 0.340 ICEPAC 0.025 – 0.325 0.093 -0.396 0.095 – 0.432 Vol52,3,2009 20-09-2009 19:06 Pagina 315 316 E.M. Warrington, A. Bourdillon, E. Benito, C. Bianchi, J.-P. Monilie, M. Muriuki, M. Pietrella, V. Rannou, H. Rothkaehl, S. Saillant, O. Sari, A. J. Stocker, E. Tulunay, Y. Tulunay and N. Y. Zaalov than for 1E, and the performance is similar for all three methods, although ICEPAC can pro- duce significantly better results in winter. Further details of this aspect of the investi- gation can be found in Pietrella et al. (2009). 5.2. Comparison of oblique sounding measurements and VOACAP predictions on a mid-latitude path between Cyprus and the UK Signals from FMCW sounders located in Cyprus were received in Leicester (a path length ~3340 km) from 1 February 2008 to 13 July 2008. The monthly median MUF as a function of time of day is presented in fig. 13. Predictions of the MUF made using VOACAP (version 05.0119W) are also presented in fig. 13. VOACAP has been run using a minimum take-off angle of 0.1°, pre- dicted smoothed sunspot numbers ranging from 7 in February to 11 in July (NGDC, 2008), and with the sporadic E model switched off. Although the observations generally have a similar form to the predictions, there are some discrepancies. For ex- ample, the predicted median MUF at 0700–0800 UT in February is much higher than the observed value (the largest daily value in the month is 21.9 MHz at 0700 UT and 23.5 MHz at 0800 UT), in June the characteristic reduction in MUF at noon is not observed, whilst in July, the observed MUF is higher than the predicted value at all times. In June and July the discrepancy results from the presence of sporadic-E layers with high (on occa- sions >30 MHz) peak frequencies. The predicted MUF(Es) values (obtained using VOACAP method 11, with a FPROB multiplier of 1.0 for Es) are closer to those observed in the morning and early afternoon, but significantly lower than the observations in the evening. Table IV. Modem availabilities with DST=0 2400 bps 1200 bps 600 bps 300 bps 150 bps 75 bps 4 MHz 0 0 0.004 0.983 0.974 1 6 MHz 0 0.026 0.245 0.891 0.869 0.983 10 MHz 0.288 0.472 0.576 0.978 0.978 0.991 11 MHz 0.179 0.424 0.533 0.987 0.974 1 14 MHz 0.122 0.258 0.380 0.987 0.978 0.996 18 MHz 0.153 0.183 0.253 1 0.991 1 Table V. Modem availabilities with DST=-050 2400 bps 1200 bps 600 bps 300 bps 150 bps 75 bps 4 MHz 0 0 0.04 1 0.96 1 6 MHz 0 0 0.12 0.96 0.92 1 10 MHz 0.32 0.4 0.56 1 0.96 1 11 MHz 0.2 0.4 0.56 1 0.92 11 14 MHz 0.08 0.2 0.28 1 0.92 1 18 MHz 0.16 0.2 0.2 1 0.96 1 Vol52,3,2009 20-09-2009 19:06 Pagina 316 317 Aspects of HF radio propagation 00 06 12 18 00 5 10 15 20 25 30 February 00 06 12 18 00 5 10 15 20 25 30 March 00 06 12 18 00 5 10 15 20 25 30 April M U F ( M H z) 00 06 12 18 00 5 10 15 20 25 30 May 00 06 12 18 00 5 10 15 20 25 30 June Time (UT) 00 06 12 18 00 5 10 15 20 25 30 July Time (UT) Fig. 13. Monthly median MUF on a path Cyprus-Leicester for February to July 2008. The observations are plotted as a solid line, while the dashed line represents predicted values from VOACAP (the predicted MUF(Es) is plotted as a dotted line in the July panel). All month-hours had at least 26 observations with the exception of July, which only had 13 for all hours. 6. Assessment of HF channel availability under ionospherically disturbed conditions Various workers have conducted investiga- tions on the signalling characteristics of HF channels, and on the influence of the channel scattering characteristics on data communica- tions capabilities. For example, Jodalen et al. (2001) reported on a comparison between mo- dem characterisations and channel measure- ments from the Doppler and Multipath Sound- ing Network (DAMSON) (Cannon et al., 2000). This system characterised narrow-band channels by measuring their SNR, Doppler spread, and delay spread parameters. Measure- ments of these characteristics may then be ap- plied to three-dimensional performance sur- faces to assess signalling capability for various types of modem. For the path from Sweden to the UK dis- cussed earlier in this paper, Warrington and Stocker (2003) previously reported the varia- tion in delay and Doppler spreading observed Vol52,3,2009 20-09-2009 19:06 Pagina 317 318 E.M. Warrington, A. Bourdillon, E. Benito, C. Bianchi, J.-P. Monilie, M. Muriuki, M. Pietrella, V. Rannou, H. Rothkaehl, S. Saillant, O. Sari, A. J. Stocker, E. Tulunay, Y. Tulunay and N. Y. Zaalov as a function of season and time of day. This previous analysis did not consider variations with ionospheric disturbance, a topic that is ad- dressed here. Joint probability density functions (PDFs) of SNR, Doppler spread, and effective multipath spread versus the Disturbance Storm Type (DST) index (see, for example fig. 14) were produced for measurements made over the path from Uppsala, Sweden (59.92º N; 17.63º E) and Leicester UK (52.63º N; 1.08º W) in 2001. It has been demonstrated by Sari (2006) by determining the conditional PDFs and by us- ing Bayes’ Theorem, that there were dependen- cies between DST and the above mentioned signal parameters. As examples of modem characterizations, various military standards were considered. In these standards, HF modem performances are specified in terms of effective multipath spread, Doppler spread, SNR, bit error rate, modulation type, and data conversion (long-short inter- leaver). Modem availability was quantified as a fraction of the time that the modem would func- tion satisfactorily (the availability of a modem can be estimated by determining the difference between the measured SNR and the SNR re- quired to give an acceptable BER for a given delay and Doppler spread - a positive SNR dif- ference indicates that the modem would operate satisfactorily (see fig. 15)). Tables IV and V show the probability of modem availability for different data rates and frequencies according to requirements defined in STANAG 4538. 7. Concluding remarks Various propagation related topics of direct relevance to the planning and operation of HF radio systems have been investigated during the Fig. 14. Joint PDF for hourly mean of SNR and negative of DST, 4 MHz (Sari, 2006). Joint pdf for hourly_mean (SNR) and neg (DST), f = 4 MHz 0.025 0.02 0.015 0.01 0.005 0 0 20 40 600 400 300 200 100 0 hourly mean (SNR) neg(DST) Vol52,3,2009 20-09-2009 19:06 Pagina 318 319 Aspects of HF radio propagation COST 296 Action, outline details of which are presented in this paper. For mid-latitude paths, these investigations have included comparisons between predicted channel MUFs for a path from Cyprus to the UK with predicted values obtained using VOACAP. Significant progress has been made in relation to the operation of HF over the horizon radars. Such systems re- quire detailed knowledge of the ionospheric electron density profile. In the work presented here, the radar is operated as a backscatter sounder for a period to measure the backscatter characteristics as a function of frequency and elevation angle. These measurements are then input to a new inversion technique to obtain the required electron density profile. The results presented in the paper illustrate the importance of understanding and taking in- to account the presence of various structural features in the northerly ionosphere, i.e. the electron depletion and irregularities associated with the sub-auroral trough, patches and arcs of enhanced electron density within the polar cap, and irregularities within the auroral zone when planning and operating HF radio links. These features result in radiowaves propagating over paths well displaced from the great circle direc- tion and have clear relevance in the operation of HF radiolocation (HF-DF) systems where devi- ations from the great circle path may result in significant (sometimes inter-continental) trian- gulation errors. A full understanding of the pre- vailing propagation mechanisms will enable ap- propriate selection of receiver sites to be made in order to optimise the positional accuracy. However, the impact of these propagation ef- fects is much wider than this particular applica- tion, extending to almost any HF communica- tions system where the signal impacts on the ionosphere within the region polewards of the Fig. 15. Approximated availability surface according to STANAG 4538 (NATO, 2000). S N R ( d B ) Available Point Unavailable Point Doppler Spread (Hz) M ultipath (m s) Vol52,3,2009 20-09-2009 19:06 Pagina 319 320 E.M. Warrington, A. Bourdillon, E. Benito, C. Bianchi, J.-P. Monilie, M. Muriuki, M. Pietrella, V. Rannou, H. Rothkaehl, S. Saillant, O. Sari, A. J. Stocker, E. Tulunay, Y. Tulunay and N. Y. Zaalov sub-auroral trough. This has been illustrated here by employing the northerly ionospheric model to indicate the effect of polar patches on the coverage area of two transmitters located at Arctic sites. Development of this aspect is cur- rently being undertaken with the aim of im- proving the HF communications forecasts and nowcasts available to the airlines when operat- ing on trans-polar routes (note that over-conser- vative forecasts of the likelihood of disturbed communications results on flights being rerout- ed to non-polar routes with consequent increas- es in flight time and fuel usage). In addition to the directionally related effects referred to above, the channel transfer characteristics are also affected by the presence of the ionospheric structures, in particular increases in delay and Doppler spread are evident. These increased spreads limit the available data throughput available on a particular channel, or alternative- ly the system availability may be reduced for a particular type of modem. Consideration of this aspect has been given for the trough path. Acknowledgements The authors would like to thank the Swedish Meteorological Institute, Uppsala and the Finnish Meteorological Institute, Helsinki for hosting the transmitter systems. The authors would also like to acknowledge support from the EPSRC under various grants. REFERENCES BATES, H.F. and R.D. HUNSUCKER (1974): Quiet and dis- turbed electron-density profiles in auroral zone iono- sphere, Radio Science, 9 (4), 455-467. BILITZA, D. (ed.) (1990): International Reference Ionosphere 1990, NSSDC 90-22, (Greenbelt, Maryland, USA). BILITZA, D. (2001): International Reference Ionosphere 2000, Radio Science, 36 (2), 261-275. BRADLEY, P.A. and J.R. DUDENEY (1973): A simple model of the vertical distribution of electron concentration in the ionosphere, Journal of Atmospheric and Terrestrial Physics, 35, 2131-2146. BUCHAU, J., B.W. RENISCH, E.J. WEBER and J.G. MOORE (1983): Structure and dynamics of the winter polar cap F region, Radio Science, 18, 995-1010. CARLSON, H.C. JR., V.B. WICKWAR, E.J. WEBER, J. BUCHAU, J.G. MOORE and W. WHITING (1984): Plas- ma characteristics of polar cap F layer arcs, Geophysi- cal Research Letters, 11, 895-898. CANNON, P.S., M.J. ANGLING and N.J. DAVIES (2000): DAMSON HF Channel Characterisation - A review, 21st Century Military Communications Conference Proceedings, 1 (2), 59-64. CCIR (1966): Atlas of ionospheric characteristics, Report 340-1, ITU, (Geneva, Switzerland). COLEMAN, C.J. (1998): A ray-tracing formulation and its application to some problems in over-the-horizon radar, Radio Science, 33 (4), 1187-1197. DYSON, P.L. and P.A. BENNETT (1988): A model of vertical distribution of the electron concentration in the iono- sphere and its application to oblique propagation stud- ies, Journal of Atmospheric and Terrestrial Physics, 50 (3), 251-262. GOLDBERG, D.E. (1989): Genetic algorithms in search, op- timization and machine learning, edited by Addison- Wesley Publishing Company. HALCROW, B.W. and J.S. NISBET (1977): A model of the F2 peak electron densities in the main trough region of the ionosphere, Radio Science, 12, 815-820. HAYDON, G.W. and D.L. LUCAS (1968): Predicting iono- sphere electron density profiles, Radio Science, 3 (13), 111-119. ITU (INTERNATIONAL TELECOMMUNICATIONS UNION) (1997): Standardized procedure for comparing predicted and observed HF sky-wave signal intensities and the pres- entation of such comparisons, ITU-R Recommenda- tion P1148-1. JODALEN, V., T. BERGSVIK, P.S. CANNON and P.C. ARTHUR (2001): Performance of HF modems on high-latitude paths using multiple frequencies, Radio Science, 36 (6), 1687-1698. JONES, R.M. and J.J. STEPHENSON (1975): A Versatile Three-Dimensional Ray Tracing Computer Program for Radio Waves in the Ionosphere, Office of Telecom- munications, OT 75-76, (U.S Department of Com- merce, Washington, USA). KIRKPATRICK, S., C.D. GELATT, Jr. and M.P. VECCHI (1983): Optimisation by Simulated Annealing, Science, 220 (4598), 671-680. LANDEAU, T., F. GAUTHIER and N. RUELLE (1997): Further improvements to the inversion of elevation-scan backscatter sounding data, Journal of Atmospheric and Solar-Terrestrial Physics, 59 (1), 125-138. NATO (2000): STANAG 4538 Technical Standards for an Automatic Radio Control System (ARCS) for HF com- munication links, (NATO Military Agency for Stan- dardisation). NGDC (2008): National Geophysical Data Centre, Boulder, Colorado USA, ftp://ftp.ngdc.noaa.gov/STP/SO- LAR_DATA/SUNSPOT_NUMBERS/2008 NORMAN, R.J., J.A. BENNETT, P.L. DYSON and J.A. NGUYEN (1994): HIRT: homing-in ray tracing pro- gram, research report, School of Physics., (La Trobe University, Bundoora, Victoria, Australia). NORMAN, R.J. and P.S. CANNON (1997): A two-dimension- al analytic ray tracing technique accommodating hori- zontal gradients, Radio Science, 32 (2), 387-396. NORMAN, R.J. and P.S. CANNON (1999): An evaluation of a new two-dimensional analytic ionospheric ray-tracing technique: Segmented method for analytic ray tracing (SMART), Radio Science, 34 (2), 489-499. Vol52,3,2009 20-09-2009 19:06 Pagina 320 321 Aspects of HF radio propagation PIETRELLA, M., E.M. WARRINGTON, A.J. STOCKER and C. BIANCHI (2009): Time of flight measurements over a radio-link from Uppsala to Bruntingthorpe and their application to testing predictions methods that approx- imate the ray tracing technique, Advances in Space Re- search, doi:10.1016/j.asr.2008.09.038. ROGERS, N.C., E.M. WARRINGTON and T.B. JONES (1997): Large HF bearing errors for propagation paths tangen- tial to the auroral oval, IEE Proceedings on Mi- crowaves, Antennas and Propagation, 144 (2), 91-96. SARI, M.O. (2006): A New Approach For The Assessment of HF Channel Availability Under Ionospheric Distur- bances, MS Thesis, Electrical and Electronics Eng., (Middle East Technical Univ., Ankara, Turkey). SIDDLE, D.R., A.J. STOCKER and E.M. WARRINGTON (2004a): The time-of-flight and direction of arrival of HF radio signals received over a path along the mid-lat- itude trough: observations, Radio Science, 39 (4), RS4008, doi:10.1029/2004RS003049. SIDDLE, D.R., A.J. STOCKER and E.M. WARRINGTON (2004b): The time-of-flight and direction of arrival of HF radio signals received over a path along the mid-lat- itude trough: theoretical considerations, Radio Science, 39 (4), RS4009, doi:10.1029/2004RS003052. STOCKER, A.J., E.M. WARRINGTON and T.B. JONES (2003): A comparison of observed and modelled deviations from the great circle direction for a 4490 km HF propagation path along the mid-latitude ionospheric trough, Radio Science, 38 (3), 1045, doi:10.1029/2002RS002781. STOCKER, A.J., N.Y. ZAALOV, E.M. WARRINGTON and D.R. SIDDLE (2009): Observations of HF propagation on a path aligned along the mid-latitude trough, Ad- vances in Space Research, 44 (1), 64-71, doi:10.1016/j.asr.2009.03.07. STEWART, F.G (undated): ICEPAC (Ionospheric Communi- cations Enhanced Profile Analysis and Circuit analysis program) Technical Manual, (available for download at http://www.greg-hand.com/manuals/icepac_tech_man- ual.pdf). TARANTOLA, A. (1987): Inverse Problem Theory, Methods for data fitting and model parameter estimation, edited by Elsevier Science. WARRINGTON, E.M. and A.J. STOCKER (2003): Measure- ments of the Doppler and multipath spread of HF sig- nals received over a path oriented along the mid-lati- tude trough, Radio Science, 38 (5), 1080, doi:10.1029/2002RS002815. WEBER, E.J., J. BUCHAU, J.G. MOORE, J.R. SHARBER, J.D. LIVINGSTON, J.D. WINNINGHAM and B.W. REINISCH (1984): F layer ionization patches in the polar cap, Journal of Geophysical Research, 89, 1683-1694. ZAALOV, N.Y., E.M. WARRINGTON and A.J. STOCKER (2003): The simulation of off-great circle HF propagation effects due to the presence of patches and arcs of enhanced elec- tron density within the polar cap ionosphere, Radio Sci- ence, 38 (3), 1052, doi:10.1029/2002RS002798. ZAALOV, N.Y., E.M. WARRINGTON and A.J. STOCKER (2005): A ray-tracing model to account for off-great circle HF propagation over northerly paths, Radio Sci- ence, 40 (4), RS4006, doi:10.1029/2004RS003183. ZOLESI, B., L.R. CANDER and G. DE FRANCESCHI (1993): Simplified ionospheric regional model for telecommu- nication application, Radio Science, 28 (4), 603-612. ZOLESI, B., L.R. CANDER and G. DE FRANCESCHI (1996): On the potential applicability of the simplified ionos- pheric regional model to different mid-latitude areas, Radio Science, 31 (3), 547-552. Vol52,3,2009 20-09-2009 19:06 Pagina 321