Vol52,3,2009 255 ANNALS OF GEOPHYSICS, VOL. 52, N. 3/4, June/August 2009 Key words Ionospheric modeling – ionospheric forecasting – ionospheric predictions 1. Introduction Ionospheric predictions of four different ty- pes are essential for mitigation of ionospheric effects on radio systems: (i) Long-term iono- spheric predictions that supply information on the ionosphere for a particular epoch of solar ac- tivity and can be used for planning radio sy- stems operations. Their success depends on the long-term solar activity prediction (Belehaki et al., 2007). (ii) Ionospheric now-casting where real- time observations are used to show the geographical and temporal variations of the io- nospheric parameters (Zolesi et al., 2008). This type of ionospheric specification is important for management of radio services in near real-ti- Near-Earth space plasma modelling and forecasting Hal J. Strangeways (1), Ivan Kutiev (2), Ljiljana R. Cander (3), Stamatis Kouris (4), Vadim Gherm (5), Diego Marin (6), Benito De La Morena (7), S. Eleri Pryse (8), Loredana Perrone (9), Marco Pietrella (9), Stanimir Stankov (10), Lukasz Tomasik (11), Ersin Tulunay (12), Yurdanur Tulunay (12), Nikolay Zernov (5) and Bruno Zolesi (9) (1) School of Electronic and Electrical Engineering, University of Leeds, Leeds, UK (2) Geophysical Institute, Bulgarian Academy of Sciences (BAS), Sofia, Bulgaria (3) Rutherford Appleton Laboratory, Didcot, UK (4) Electrical and Computer Engineering Department, Aristotle University of Thessaloniki, Greece (5) Department of Radiophisics, University of St. Petersburg, Russian Federation (6) University of Huelva, Huelva, Spain (7) Atmospheric Sounding Station El Arenosillo, INTA, Huelva, Spain (8) Aberystwyth University, Aberystwyth, UK (9) Istituto Nazionale di Geofisica e Vulcanologia, Roma, Italy (10) Royal Meteorological Institute, Brussels, Belgium (11) Center for Space Research, Warsaw, Poland (12) Middle East Technical University (METU), Ankara, Turkey Abstract In the frame of the European COST 296 project (Mitigation of Ionospheric Effects on Radio Systems, MIERS) in the Working Package 1.3, new ionospheric models, prediction and forecasting methods and programs as well as ionospheric imaging techniques have been developed. They include (i) topside ionosphere and meso-scale ir- regularity models, (ii) improved forecasting methods for real time forecasting and for prediction of foF2, M(3000)F2, MUF and TECs, including the use of new techniques such as Neurofuzzy, Nearest Neighbour, Ca- scade Modelling and Genetic Programming and (iii) improved dynamic high latitude ionosphere models through tomographic imaging and model validation. The success of the prediction algorithms and their improvement over existing methods has been demonstrated by comparing predictions with later real data. The collaboration between different European partners (including interchange of data) has played a significant part in the development and validation of these new prediction and forecasting methods, programs and algorithms which can be applied to a variety of practical applications leading to improved mitigation of ionosphereic and space weather effects. Mailing address: Dr. Hal J. Strangeways, University of Leeds, Leeds, UK; e-mail: H.J.Strangeways@leeds.ac.uk Vol52,3,2009 22-09-2009 14:19 Pagina 255 256 H.J. Strangeways, I. Kutiev, L.R. Cander, S. Kouris, V. Gherm, D. Marin, B. De La Morena, S. E. Pryse, L. Perrone, M. Pietrella, S. Stankov, L. Tomasik, E. Tulunay, Y. Tulunay, N. Zernov and B. Zolesi me. (iii) Shortterm ionospheric prediction that can be based on extrapolation of past data sets, forecasting of a short-term disturbance index on which the ionosphere critically depends and mo- nitoring of ionospheric propagation conditions in connection with solar-terrestrial parameters (Bremer et al., 2006). This is required to impro- ve the quality and reliability of radiocommuni- cation services, including frequency adaptive applications at MF and HF and transionospheric radio determination (Wheadon et al., 2006). (iv) Prediction of special disturbing factors like scin- tillations, which disturb GPS signals. The terms of reference of Working Package 1.3 (WP 1.3) «Near-Earth space plasma model- ling and forecasting» comprised a large area of near-Earth plasma modelling, supporting basic research, forecasting and radio communica- tions. The structure of this paper follows the main topics of investigations in WP 1.3, viz: 2. Topside model development and testing. 3. De- velopment of techniques for real-time foreca- sting. 4. Forecasting of foF2, M(3000)F2, MUF and TEC. 5. Tomographic imaging and model validation. 6. Modelling mid-scale (meso-scale) inhomogeneities at low-latitude (bubbles) and high-latitude (patches, blobs) to predict their scintillation effects. A catalogue of ionospheric forecasting and predicting models developed in the framework of COST actions 238, 251, 271 and 296 has been compiled. The catalogue and brief de- scription of all 18 models are available at: http://www.cost296.rl.ac.uk/word/Catalogue- Text.doc 2. Topside models development and testing 2.1. Topside electron density model Triskova et al. (2007), have developed a model, providing electron density distribution at fixed altitudes as functions of invariant latitu- de and magnetic local time. A system of asso- ciated Legendre polynomials up to the 6th-or- der are fitted to the data from four satellites: Hi- notori, ISS B, ISIS 1, and ISIS 2. The full mo- del consists of submodels for individual sea- sons and altitude ranges. Triskova et al., 2007 have developed a topside electron density mo- del by using vary-Chap functions for the scale height. At each given location, Ne submodels provide the anchor points for each altitude le- vel, which are then used to construct the verti- cal profiles. This new topside presentation avoids the «kinks» in the profiles resulting from applying the Booker formulism. An example of the topside model is shown in fig. 1. It is one of options in the International Reference Iono- sphere (IRI). 2.2. TSMP-assisted topside reconstruction To improve the accuracy of the real time to- pside electron density profiles given by the Di- gisonde software, a new model-assisted techni- que is used (Kutiev et al., 2009). This technique uses the Topside Sounder Model (TSM), which provides the plasma scale height (Hs), O+-H+ transition height (HT), and their ratio Rt=Hs/HT, derived from topside sounder data of Alouette and ISIS satellites. Kutiev and Marinov (2007) introduced the concept of a new profiler «TSMP», using their scale height (Hs) and tran- sition height (HT) models. TSMP provides the shape of the topside N(h) profile as the sum of the O+and H+profiles from hmF2 up to GPS hei- ghts. The O+ profile takes Hs as its own scale height (HO+=Hs), at the transition height both densities are equal, and the scale height above H+ (HH +) is taken as 16 times the O+ scale hei- ght. For the present analysis we use TSMP in the form: (2.1) F2(hT)=NO+=NH+ at transition height hT. To cal- culate the topside density profile, TSMP needs values of NmF2 and hmF2 at the lower boun- dary (denoted as Nm and hm), which are sup- plied by Digisonde measurements. In addition, the Digisonde also provides the topside scale 2 ,exp exp exp exp N N H h h H h h N N H h h Ne F h F h F h H h h 2 1 1 2 2 m m m H H m H T T H T 0 0 0 = - - - - - = - - = = + - - + + + + + + + b ] b ] ] b l g l g g l ; E' 1 Vol52,3,2009 20-09-2009 19:06 Pagina 256 257 Near-Earth space plasma modelling and forecasting height Hd. It is assumed that, for a particular moment in time, Hd represents the real condi- tions, and is therefore a better value than the average model value of Hs. When Hd is speci- fied, the corresponding transition height hT is obtained through the ratio Rt, e.g. hT = Hd.Rt. To illustrate the general concept, we apply TSMP to reconstruct the topside Ne profile in two particular cases. Figure 2 show two exam- ples of profile reconstruction for Athens. 2.3. TSM model development and testing The topside ionospheric scale height (Hs) and the O+-H+ transition height (HT) are key io- nospheric parameters which are of special inte- rest when studying and modelling the plasma composition and dynamics. Recently, the Hs (Kutiev et al., 2006) and HT (Marinov et al., 2004) empirical models have been combined in a unified Topside Sounder Model (Kutiev and Marinov, 2007). The database for these models has been built upon 170,000 Hs and HT values deduced from electron profiles obtained via to- pside sounding measurements. For validation purposes, Stankov et al. (2007) compared TSM model results with those from the well-known models NeQuick and the Parameterized Iono- spheric Model (PIM). Electron density profi- les, in the altitude range 200-2000 km, are ob- tained from the NeQuick and PIM models for a grid of input parameters, such as month, local time, geomagnetic latitude, and solar activity. The scale height and transition height were ex- tracted from the model profiles in the same way they were extracted from the measured topside sounder profiles. Their values are then compa- red with the respective values provided by the TSM model. Comparison of the scale height and the transition height show that, in general, the NeQuick and PIM models provide higher values of both scale height and transition height than those of TSM. Fig. 1. Examples of Ne vertical profiles constructed by using densities at the sub-model altitudes, taken as an- chor values, along with corresponding IRI models, for two levels of solar activity. Vol52,3,2009 20-09-2009 19:06 Pagina 257 258 H.J. Strangeways, I. Kutiev, L.R. Cander, S. Kouris, V. Gherm, D. Marin, B. De La Morena, S. E. Pryse, L. Perrone, M. Pietrella, S. Stankov, L. Tomasik, E. Tulunay, Y. Tulunay, N. Zernov and B. Zolesi Briefly, the comparison results for the pla- sma scale height can be summarized as follows: • PIM - generally highest values during hi- gh solar activity • PIM - unusually high values during an HSA day at low and middle latitudes • NeQuick - highest values (unusually hi- gh) in LSA winter. The comparison results for the ion transition height can be summarized as follows: • PIM - systematically higher (probably overestimated) values at HSA • NeQuick - relatively weaker dependence on solar activity • NeQuick - discontinuities over the equa- tor occur at night time. 2.4. Modeling TEC and slab thickness variations Considerable efforts have been devoted to quantifying the response of TEC to changes in solar activity, location, season and time of day. Kouris et al. (2006; 2008) found that TEC is hi- ghly correlated with solar activity and latitude and also with season. Applying a least squares analysis to the vertical TEC obtained from GPS measurements at different European locations and the corresponding solar flux density data at 10.7 cm, the total electron content relative to the quiet sun level (To ) at each hour/ month/ loca- tion is counted. Then, these To values are corre- lated with corresponding values of the seasonal parameter cosχ at noon at each location accor- ding to the following least squares equation To = k (cosχ12 )m (2.2) where cosχ12 is the cosine of the solar zenith an- gle at noon (a seasonal parameter). The calcula- ted values of the factor ‘k’ and the exponent ‘m’ determined by using monthly median vertical TEC values measured at noon and/or any other hour, exhibit a latitudinal dependence (Kouris et al., 2009) since TEC depends on latitude. The slab thickness, an important parameter in the determination of N2max from TEC measu- rements, is also investigated and some specifica- tions are established. It has a diurnal variation (Kouris et al., 2006) in winter from about 200 km to 450 km on average, whereas in equinoxes Fig. 2. TMSP profile reconstruction for Athens: at 04:30 UT (left) and 16:30 UT (right) on 1 October 2000. TSMP-assisted Digisonde (TaD) profiles are shown the heavy line; solid and dashed curves show O+ and H+ dis- tributions respectively. For comparison, CHAMP-based reconstruction (Heise et al., 2002) profiles are given by the grey curves. Vol52,3,2009 20-09-2009 19:06 Pagina 258 259 Near-Earth space plasma modelling and forecasting and summer, this ranges from circa 300 km to 450 km depending on the time of day. In winter and equinoxes the nighttime values are much hi- gher than the day-time, whereas in summer the opposite occurs so that the day values are higher than the nighttime ones. The predawn peak is found to occur in winter and equinoxes from about 500 km to 700 km and in summer from circa 400 km to 450 km. The monthly average values of the slab thickness all over Europe are practically the same, within a range of circa 20 km to 30 km during day-time. Some greater dif- ferences exist during nighttime due to disturbed conditions which depend on location and occur predominantly during nighttime (Fotiadis et al., 2004; Kouris et al., 2005). The equivalent slab thickness is independent of solar activity and la- titude (Kouris et al., 2008) but depends strongly on season (Kouris et al., 2009). A mean value of the equivalent slab thick- ness for day-time conditions (cosine of the so- lar zenith angle greater or equal to 0.10) can be determined from the following analytical ex- pression: Est. = 370 (cosχ12 ) 0.44 (2.3) For nighttime conditions (i.e. cosine of the solar zenith angle less than 0.10) a mean value can be calculated from the equation: Est = 313 (cosχ12 ) -0.24 (2.4) These relationships are valid for locations within the latitudinal range of 30 to circa 60 de- grees. Higher latitude locations are excluded because of the decreased stability of the F-re- gion in this zone (Rawer et al., 2003). Neglec- ting diurnal and seasonal variations, an overall mean slab thickness for Europe (circa 30oN to 55oN) can be taken as 323 km with a standard deviation less than 50 km. 3. Development of techniques for real-time forecasting A study has been carried out on those major areas where our current physical understan- ding and recent advances can lead to positi- ve prediction, forecasting and mitigation of the expected effects of ionospheric activity on the near Earth space environment and on technolo- gical systems which operate within this envi- ronment (Cander and Mihajlovic, 2005; Cop et al., 2008; Cander and Ciraolo, 2008). The results show some of the key links between solar activity and the various physical processes, which govern ionospheric plasma structure. Specific examples during extremely intense solar event as well as during low solar activity illustrate how ionospheric monitoring techniques that have contributed immense data sets and related empirical and theoretical for- mulations can be incorporated in different mo- dels for real-time operational applications (Jan- sen et al., 2006; Cander 2009). An example of the long-running services developed in the course of the European iono- spheric COST actions is the «Space Weather Web Facilities for Radio Communications Users» at Rutherford Appleton Laboratory, UK since 1998. In this on-line service the real-time ionospheric observations have been integrated into suitable methods and models to provide a number of products. An overall extensive vali- dation procedure for these products has shown that the combined subsequent ionospheric pre- dictions are essential for any mitigation of iono- spheric effects on radio systems: 1) Long-term ionospheric predictions that supply information on the ionosphere for a particular epoch of so- lar activity and can be used for planning radio systems operations. Their success depends on the long-term solar activity prediction (Be- lehaki et al., 2007); 2) Ionospheric now-casting where real-time observations are used to show the geographical and temporal variations of the ionospheric parameters (Zolesi et al., 2008). This type of ionospheric specification is impor- tant for management of radio services in near real-time; and 3) Shortterm ionospheric foreca- sting (STIF) that can be based on extrapolation of past data sets, forecasting of a short-term di- sturbance index on which the ionosphere criti- cally depends and monitoring of ionospheric propagation conditions in connection with so- lar-terrestrial parameters (Muhtarov et al., 2002; Bremer et al., 2006). Figure 3 shows a comparison between the Vol52,3,2009 20-09-2009 19:06 Pagina 259 260 H.J. Strangeways, I. Kutiev, L.R. Cander, S. Kouris, V. Gherm, D. Marin, B. De La Morena, S. E. Pryse, L. Perrone, M. Pietrella, S. Stankov, L. Tomasik, E. Tulunay, Y. Tulunay, N. Zernov and B. Zolesi foF2 produced by the short-term ionospheric forecasting method and the data from Chilton (51.50 N, -1.30 E) ionosonde. It can be seen that this forecasting facility performs very well du- ring quiet geomagnetic periods. Although it fal- ters at intense geomagnetic events, it still provi- des a valuable ionospheric predictor over Euro- pe, having in mind the storm distribution given in fig. 4. 4. Forecasting of FoF2, M(3000)F2, MUF and TEC 4.1. foF2 forecast during severe geomagnetic activity in Rome observatory To forecast the ionospheric response to geo- magnetic storms, geomagnetic indices have been introduced taking into account their history (Wrenn, 1987; Wu and Wilkinson, 1995; Perro- ne and De Franceschi, 1999). One of these is ap(τ) (Wrenn, 1987) derived with a time weigh- ted series accumulation from the geomagnetic planetary index ap. The attenuation multiplier τ ranges between 0 and 1 and determines the wei- ghts assigned to past ap values. The larger is τ, the smoother are the ap variations. An impro- ved linear correlation was found between tran- sformed data obtained from hourly and monthly median values of foF2 and ap(τ) with τ= 0.8 (Perrone and De Franceschi, 1999). The tran- sformed data of foF2 that gives the best result is log(NmF2(t)/NmF2M(t)), where NmF2(t) is the hourly value of the maximum electron density at the F2 peak and the suffix M indicates the monthly median value (Wu and Wilkinson, 1995). Instead for the monthly median repre- senting the «quiet» ionosphere, an average is calculated by Perrone et al. (2007). It is obtai- ned considering, for a selected hour, foF2 va- lues with ap(τ)�7 in the thirty days preceding the day for which we want to have the «quiet» value. This average is called the «daily» mean. Fig. 3. Measured and 24 hours in advance forecast foF2 values during quiet and disturbed days in January 1999 at Chilton ionospheric station. Vol52,3,2009 20-09-2009 19:06 Pagina 260 261 Near-Earth space plasma modelling and forecasting The data utilized are the foF2 hourly values measured at the Rome (41.9° N; 12.5° E) obser- vatory. Geomagnetic storms with a maximum of ap�132 are classified as strong events. An example of foF2, forecast with the rela- tionship between the foF2 transformed data and ap(τ), is presented in fig. 5, which illustrate the results of forecasting for the period 12-17 July 2000, when a negative ionospheric storm occur- red. The r.m.s. of forecast is 1.23 MHz and for the «daily mean» it is 2.0 MHz. The model was further developed to adapt it for different seasons and ap levels. The new mo- del, called IFELMOR (Ionospheric Forecasting Empirical Local Model Over Rome, Pietrella and Peronne, 2008) uses different sets of coeffi- cients for each month of the year and several ranges of ap(τ). IFELMOR predictions are com- pared in table I with those of the IRI-STORM model in terms of r.m.s. error for selected days with very disturbed ionospheric conditions (daily mean of ap(t = 0.9), ap(τ = 0.9) > 70). The best performance is labelled in bold. 4.2. Neurofuzzy techniques applied to model and predict the F2-layer critical frequency foF2 A new methodology to predict the iono- spheric F2-layer critical frequency, foF2, 1-24 hours in advance has been advanced. The pro- posed method is based on artificial intelligence techniques, specifically, on neuro-fuzzy model- ling, very little used in ionospheric modelling until now. Such techniques have a natural capa- bility to model well non-linear and complex sy- stems. Neurofuzzy models incorporate from fuzzy models the ability to understand and con- nection with physical processes and, on the other hand, from neural networks based mo- dels, the capability of adaptation and learning. Neurofuzzy models allow incorporation of ex- Fig. 4. Years count of magnetic storms during solar cycle 23 (1996-2007). 14 12 10 8 6 4 2 0 1996 1997 1998 1999 2000 2001 Years N u m b e r o f st o rm s w ith D st < -1 0 0 n T 2002 2003 2004 2005 2006 2007 Vol52,3,2009 20-09-2009 19:06 Pagina 261 262 H.J. Strangeways, I. Kutiev, L.R. Cander, S. Kouris, V. Gherm, D. Marin, B. De La Morena, S. E. Pryse, L. Perrone, M. Pietrella, S. Stankov, L. Tomasik, E. Tulunay, Y. Tulunay, N. Zernov and B. Zolesi pert knowledge in different parts of the model- ling process. To predict foF2 n hours ahead, the values of foF2 at the current time t, at two previous hours, and 24 and 25 hours in the past are considered, along with the respective ap values. For predic- tions at 1 or 2 hours ahead, foF2 values from the previous day are not taken into account. In order to test the foF2 neurofuzzy models, the relative mean deviation (RMD) between ob- servations and short-term foF2 predictions (1-24 hour lead times) were calculated for 100 periods of 3 consecutive days. Values of foF2 were ob- served at Slough in 1980-82. 62.1% of the selec- ted 300 days are geomagnetic quiet days (Ap=10), the other 37.9% correspond to mode- rate geomagnetic activity (1040. Despite this poor accuracy for lead times larger than 3 hours, the great efficiency of the neurofuzzy model to pre- dict 1-3 hours in advance, even under very di- sturbed periods, should be noted. To illustrate this result, the variations of observed and 1- hour predicted foF2 values for three time pe- riods, when effects of strong negative storms were observed, are represented in fig. 6. It can be noticed how both observed and predicted foF2 variations are very close. Table I. Comparison of predictions between IFELMOR and IRI-STORM. Day/month/year IFELMOR STORMMEDIANS MEDIANS STORMfoF2QT foF2QT ap(t = 0.9) r.m.s (MHz) r.m.s (MHz) r.m.s (MHz) r.m.s (MHz) r.m.s (MHz) 16th July 2000 0.47 0.56 1.90 0.42 2.24 77.8 31th March 2001 1.70 2.81 2.97 4.11 4.48 77.1 1th April 2001 1.54 1.71 1.98 2.02 3.34 78.7 30th October 2003 0.87 1.87 1.89 6.78 6.83 125.1 31th October 2003 0.83 2.33 2.34 6.62 6.65 147.8 1th November 2003 1.88 0.83 0.73 3.03 2.83 81.6 21th November 2003 0.73 0.76 0.80 1.55 1.40 77.8 Global r.m.s. error 1.34 1.78 1.94 4.02 4.25 Table II. Relative mean and standard deviations between observations and predictions. Lead Time (hours) RMD(SD) 1 2 3 6 12 18 24 3.92 7.91 8.24 9.11 9.73 9.88 9.92 (1.65) (1.36) (1.45) (1.51) (1.83) (2.06) (1.85) Vol52,3,2009 20-09-2009 19:06 Pagina 262 a p (τ ) 263 Near-Earth space plasma modelling and forecasting Fig. 5. The behaviour of foF2 observed, forecasted and of the «daily» mean for 12-17 July 2000. Fig. 6. 1-hour prediction examples over 3 very disturbed periods. The Ap index value is given to indicate the geomagnetic activity level. 0 24 48 72 96 120 2 4 6 8 10 12 14 16 Observations 1-hour Predictions RMD (%) = 5.30 fo F 2 V a lu e s (M H z) 0 24 48 72 96 120 144 19.4 61.9 42.9 13.1 18.8 7.3 25 26 27 28 29 30 Ap Value: 39.0 95.8 120.6 26.8 8.1 Day: 11 12 13 14 15 Observations 1-hour Predictions RMD (%) = 3.84 May 1981 Time (hours) April 1981 April 1981 0 24 48 72 96 120 11.3 39.4 32.0 60.8 13.3 8 9 10 11 12 Observations 1-hour Predictions RMD (%) = 4.36 fo F 2 (M H z) 12-17 JULY 2000 200 150 100 50 0 12 10 8 6 4 2 12 13 14 15 16 17 DAYS Vol52,3,2009 20-09-2009 19:06 Pagina 263 264 H.J. Strangeways, I. Kutiev, L.R. Cander, S. Kouris, V. Gherm, D. Marin, B. De La Morena, S. E. Pryse, L. Perrone, M. Pietrella, S. Stankov, L. Tomasik, E. Tulunay, Y. Tulunay, N. Zernov and B. Zolesi 4.3. Forecasting of foF2 and M(3000)f2 using the Nearest Neighbour (NN) algorithm Forecasting of ionospheric characteristics has been accomplished by means of the nearest neighbour and modified neural network algo- rithms. Algorithms were used for data gaps cor- rection, as well as the characteristics’ forecast. They are determined for the European area. Both models are in operational use at RWC Warsaw (Space Research Centre). A full de- scription is given on the COST 296 WG1 home page (http://rwc.cbk.waw.pl/cost296/wg1/) (Io- nosphere forecasting models). Table III illustra- tes the quality of forecast through the RMS er- ror for prediction 24 hours ahead. 4.4. Neural Network based Forecasting, Maps, and Process Identification The Middle East Technical University (ME- TU) data driven models have proven to be power- full in forecasting the parameters of the non-li- near processes including the ionospheric proces- ses. These models consist of Neural Networks (NN), Neuro Fuzzy Networks (NFN), Genetic Programming (GP) and Cascade Models. The models are capable of performing now- casting and forecasting of ionospheric parame- ters such as ionospheric critical frequancy (foF2), Total Electron Content (TEC) values up to 24 hours in advance for single and multista- tions and also performing forecast mapping over the COST 296 area during disturbed Spa- ce Weather (SW) conditions (Tulunay et al., 2004; Tulunay et al., 2006; Senalp et al., 2008). Some illustrative examples for mapping ap- plication are presented. In the first example, the METUNN-C model has been employed to fore- cast the GPS TEC grid values over Europe (Se- nalp, 2007). Bezier surfaces are used for mapping including the SW events in November 2003 who- se TEC Absolute Error Map is presented in fig. 7. In the second example, as illustrated in fig. 8, the forecast foF2 values are obtained by us- ing the METU Fuzzy-Neural Network model (METU-FNN) (Altuntas, 2007) during the Hal- loween 2003 storm and the mapping is per- formed using Genetic Programming (GETY) (Yapici, 2007). 4.5. Tests of MUF forecasts In order to test the preliminary results obtai- ned in a previous research (Zolesi et al., 2008), the hourly MUF measurements were compared with the hourly MUF predicted by the longterm models ASAPS, ICEPAC and SIRM&LKW and with the hourly values of MUF predicted by the nowcasting models SIRMUP&LKW and ISWIRM&LKW. The comparisons in terms of the global r.m.s. error between the long term models and the nowcasting models, carried out for the 33 days selected among different sea- sons and geomagnetic conditions, quantified in table IV how much better is the nowcasting pre- Table III. RMS error for prediction 24 hours ahead. Name RMS foF2 [MHz] RMS M3000f2 El Arenosillo 1.15 0.2 Juliusruh 0.90 0.17 Loparskaya 1,2 x Sofia 0.90 0.16 Tashkent 1,70 x Tortosa 0.98 0.17 Vol52,3,2009 20-09-2009 19:06 Pagina 264 265 Near-Earth space plasma modelling and forecasting diction of individual days than the longterm model prediction during winter months and equinoctial months, as well as under moderate, disturbed, and very disturbed geomagnetic con- ditions (Pietrella et al., 2009). 5. Tomographic imaging and model validation During the evening and nighttime the mid- latitude ionosphere is separated from the auro- ral region by the main ionisation trough (Pryse et al., 2006a). The mechanisms that form and sustain the trough have been studied for many years but are still open to debate, in particular in the post magnetic midnight sector. Recent stu- dies by the Solar System Physics research group at Aberystwyth University have shown the role of plasma drawn across the polar region in sustaining the poleward wall of the trough (Pryse et al., 2007). Radiotomography observa- tions by the International Ionospheric Tomo- graphy Community (IITC) have been the focus of the work, reinforced by measurements from other experimental techniques and the Coupled Thermosphere-Ionosphere-Plasmasphere (CTIP) model. In turn, the observations have served to validate and develop the model. Several studies have investigated the role that plasma transported from the polar region plays in the nightside ionosphere. One example using observations by radiotomography and the EISCAT and SuperDARN radars concluded that in this instance enhanced ionisation was drawn from the polar region, through the Ha- rang discontinuity and into the pre-midnight sector where it was restructured into a boundary blob (Pryse et al. 2006b). Radiotomography ob- servations in Scandinavia and UK (Middleton Fig. 7. Absolute error map for observed and 1 h ahead forecast TEC during 16-29 Nov. 2003. A b so lu te E rr o r V a lu e ( T E C U ) Latitude (°) Longitude (°) Vol52,3,2009 20-09-2009 19:06 Pagina 265 266 H.J. Strangeways, I. Kutiev, L.R. Cander, S. Kouris, V. Gherm, D. Marin, B. De La Morena, S. E. Pryse, L. Perrone, M. Pietrella, S. Stankov, L. Tomasik, E. Tulunay, Y. Tulunay, N. Zernov and B. Zolesi Fig. 8. Superimposed are the observed and 1-h in advance forecast values of foF2 for Sofia between 12 and 24 Nov. 2003. Table IV. Comparison of RMS errors between nowcasting and long term models. ASAPS SIRM ICEPAC SIRMUP&LKW ISWIRM&LKW ap daily mean r.m.s. (MHz) r.m.s. (MHz) Prof CCIR r.m.s. (MHz) r.m.s. (MHz) r.m.s. (MHz) 23 June 2005 3.21 2.89 3.36 2.29 1.61 50 10 July 2005 3.26 2.59 3.33 2.18 1.52 57 12 September 2005 2.51 2.70 2.50 1.55 2.32 75 18 January 2005 3.48 4.06 2.79 2.52 3.11 84 15 May 2005 3.30 2.80 3.18 1.82 0.97 87 8 May 2005 3.77 3.28 3.70 2.05 0.98 91 24 August 2005 3.83 3.22 3.99 2.42 2.65 102 Global r.m.s. error 3.13 3.35 3.26 2.12 2.03 fo F 2 (M H z) Super Storm time period (12/11/2003 - 24/11/2003) (hour) Vol52,3,2009 20-09-2009 19:06 Pagina 266 267 Near-Earth space plasma modelling and forecasting et al., 2008) under IMF Bz negative revealed a prominent poleward trough wall in the post ma- gnetic midnight sector. Using the CTIP model and multi-instrument measurements the ionisa- tion of this wall was interpreted as arising from a tongue-of-ionisation (TOI) drawn anti- sunward through the polar cap, through the Ha- rang discontinuity and into the dawn sector where it formed the trough wall. Other studies have shown dayside ionisation drawn antisunward towards the nightside under conditions of IMF Bz positive. An example using radiotomography measurements from the Scandinavian and Greenland sectors (Middleton et al., 2005) revealed enhanced density on the periphery of the polar cap in the afternoon and evening. This was interpreted as a TOI being drawn around lobe cells in the polar cap, rather than being drawn directly across the polar re- gion. Wood et al. (2008) investigated the spatial variation in the nighttime ionisation under Bz positive, and revealed enhanced ionisation on the nightside under Bz positive, but with a re- duction in density in a region where it was likely that the lobe cells were blocking the transit of ionisation from the dayside to the nightside. Initial comparisons of the CTIP model out- put with observations (Middleton et al., 2008) showed disparities in the detail of the location and density of the poleward trough wall. In a subsequent study, electric potential patterns shaped by SuperDARN plasma flow observa- tions were used to replace library patterns as in- put to the model. A significant improvement was obtained between the observed and model- led enhancements forming this wall (Pryse et al., 2009; Whittick et al., 2009). Sample comparisons between the model and observations are shown in fig. 9 under con- ditions of IMF Bz negative, where the density enhancement in the equatorward field-of-view of the panels represents a cross-section through a TOI swept into the post-magnetic midnight sector to form the poleward trough wall. For conditions of IMF Bz positive, the CTIP model with a SuperDARN electric potential pattern yielded dayside ionisation drawn anti- sunward around the periphery of the polar re- gion, in accord with tomography observations under IMF Bz positive (Pryse et al., 2009). The library of electric potential patterns, used with CTIP prior to the use of SuperDARN observa- tions, did not include lobe cells characteristic of IMF Bz positive required to reproduce a spatial plasma distribution akin to that observed. 6. Modelling mid-scale (meso-scale) inhomogeneities at low-latitude (bubbles) and hight-latitude (patches, blobs) for scintillation predictions An analytical model has been developed for mid-scale (meso-scale) ionospheric inhomoge- neities in the background ionosphere occurring either as patches in the high-latitude ionosphe- re or bubbles in the low latitude/equatorial io- nosphere. The mid-scale irregularity model is embed- ded in the large-scale background ionosphere in order to investigate and predict their scintilla- tion effects. Separate models were constructed for both ionospheric electron density bite-outs (bubbles) and enhancement (patches) together with the smaller rod-like irregularities contai- ned within them. To construct the mid-scale inhomogeneity model, a basic function was em- ployed of the form: (6.1) The pattern given by function (6.1) is shown in fig. 10. The 3-dimensional structure can be modelled by means of functions of type (6.1) as the product:- (6.2) (6.3) (6.4) In the approximation of the «frozen drift» the positions of the «walls» are functions of time whereas the shape of a structure is kept the same. , , , , ,v a b c v a b c1 1 1 1 2 2 2 2= =" ", , , , , , , , , ,r x y z r x y z r x y z1 1 1 1 2 2 2 2= = =" " ", , , , , , , . , , , , , , , , . , , , , r r v r v F y y b y b F x x a x a F z z c z c 1 1 2 2 1 1 2 2 1 1 2 2 1 1 2 2 Φ =^ ^ ^ ^ h h h h , , , , , . exp exp F x x a x a a x x a x x x x 1 1 < 1 1 2 2 1 1 1 2 2 1 1 2 = + - - - - + - - - - ^ b b h l l ; ; E E Vol52,3,2009 20-09-2009 19:06 Pagina 267 268 H.J. Strangeways, I. Kutiev, L.R. Cander, S. Kouris, V. Gherm, D. Marin, B. De La Morena, S. E. Pryse, L. Perrone, M. Pietrella, S. Stankov, L. Tomasik, E. Tulunay, Y. Tulunay, N. Zernov and B. Zolesi Embedding the mid-scale inhomogeneity into the ionosphere is managed by modulation of the background ionosphere utilizing equation 6.2: (6.5) A is a constant, which is positive for an enhanced level of the electron density in a local structure and negative for a depletion. To model the elec- tron density fluctuations, a fully three-dimensio- nal inverse power law spatial spectrum of the cor- , , , , ,Ne r t Ne r A r r t v r t v1 10 1 2 2$ $ Φ= +^ ] ] ]^h g g g h7 A relation function was formulated, including an aspect ratio for two mutually perpendicular tran- sverse (to the geomagnetic field) directions as well as a longitudinal to transverse one. The mo- del of the electron density fluctuations is given by: (6.6) α is the along-magnetic field and ß the across- , / / , k R p p R L 3 2 2 1 2 /N N x y z p 0 3 3 2 2 0 2 2 2 2 2 2 2 0 κ π αβ σ κ κ β κ α κ κ π Ψ Γ Γ = - + + + = - ^ ^^ ^ ] d h h h h g n Fig. 9. Plasma distributions modelled by the CTIP model (left-hand-side panels) at the 18ºE meridian in the northern hemisphere at 00UT and 03UT on 13 December 2001, and corresponding radio tomography observa- tions (right-hand-side panels). The contours are in units of 1011m-3. Vol52,3,2009 20-09-2009 19:06 Pagina 268 269 Near-Earth space plasma modelling and forecasting magnetic field aspect ratio. where L is the smallest outer scale of the fluctuations. The variances of the fractional electron density fluc- tuations along the ray paths are given by; (6.7) Details of the propagation model, calcula- tion of the two-point correlation and cross cor- relation functions of the phase and log-amplitu- de of the field, estimation of the statistical cha- racteristics of the received field and resultant simulated time series of received amplitude and phase can be found in (Gherm et al., 2005; Maurits et al., 2008). The parameters of the spectrum of the elec- tron density fluctuations, as well as the meso-sca- le local polar structures (e.g. bubbles or patches) are chosen empirically. The NeQuick model [http://www.itu.int/ITU-R/study-groups/softwa- , , , , , ,r t r t B r r t v r t v1fr fr 2 0 2 1 1 2 2$ $σ σ Φ= +^ ^ ] ]^h h g g h7 A 2 k L0 π = re/rsg3-p531-electron-density.zip] was used to describe the low latitude/equatorial ionosphere. (Gherm et al., 2007; Strangeways et al., 2007) describe using the meso-scale model to predict scintillation effects for bubbles and the good agreement obtained with experimental data taken at Douala, Cameron (Maurits et al., 2008; Zernov et al., 2008; 2009) describe the application to mo- delling propagation through polar patches. The University of Alaska, Fairbanks Eulerian Parallel Polar Ionosphere Model was used to re- present the large-scale background polar iono- sphere. The resultant propagation simulator enabled realistic predictions of the scintillation effects for high-latitude transionospheric propagation at L- band (e.g. GPS signals) in the presence of polar patches for a number of different propagation scenarios for L-band transionospheric signals. For some more restricted conditions, the propa- gation model can also produce results for lower frequencies, such as 400 MHz, where the regime of strong scintillation arises. Fig. 10. Plasma bubble/polar patch model. 1 0.8 0.6 0.4 0.2 0 0 2 4 6 8 10 x1=2 a1=0.25 x2=7 a2=0.5 F (x ) x Vol52,3,2009 22-09-2009 14:52 Pagina 269 270 H.J. Strangeways, I. Kutiev, L.R. Cander, S. Kouris, V. Gherm, D. Marin, B. De La Morena, S. E. Pryse, L. Perrone, M. Pietrella, S. Stankov, L. Tomasik, E. Tulunay, Y. Tulunay, N. Zernov and B. Zolesi REFERENCES ALTUNTAS, E. (2007): Forecasting of the EM Waves in Ionised Media Related to Aerospace Applications, MS Thesis, Aerospace Eng., Middle East Technical Univ., (Ankara, Turkey, Sep. 2007). BELEHAKI, A., LJ.R. CANDER, B. ZOLESI, J. BREMER, C. JUREN, I. STANISLAWSKA, D. DIALETIS and M. HATZOPOULOS (2007): Ionospheric specification and forecasting based on observations from European ionosondes participating in DIAS project, Acta Geophysica, 55 (3), 398-409. BREMER, J., LJ.R. CANDER, J. MIELICH and R. STAMPER (2006): Derivation and test of ionospheric activity in- dices from real-time ionosonde observations in the Eu- ropean region, Journal of Atmospheric and Solar-Ter- restrial Physics, 68 (18), 2075-2090. CANDER, LJ.R and S.J. MIHAJLOVIC (2005): Ionospheric spatial and temporal variations during the 29-31 Octo- ber 2003 storm, Journal of Atmospheric and Solar-Ter- restrial Physics, 67 (12), 1118-1128. CANDER, LJ.R. and L. CIRAOLO (2008): Total Electron Con- tent over Europe at Solar Minimum: Storm Case, 1st GNSS Vulnerabilities and Solutions Conference, (Baška, Croatia, 7-9 September). CANDER, LJ.R. (2009): Ionospheric ground-based measure- ment networks, in Proceedings of the International He- liophysical Year, Earth, Moon and Planets, 104, 37-40, doi:10.1007/s11038-008-9245-z. COP, R., S.J. MIHAJLOVIC and LJ.R CANDER (2008): Mag- netic Storms and their Influence on Navigation, Po- morstvo, Journal of Maritime Studies, 22 (1), 89-99. FOTIADIS, D.N., S.S. KOURIS, V. ROMANO and B. ZOLESI (2004): Climatology of ionospheric F-region distur- bances, Annals of Geophysics, 47 (5), 1311-1323. GHERM, V.E., N.N. ZERNOV and H.J. STRANGEWAYS (2005): Propagation model for transionospheric fluctuating paths of propagation in a wideband ionospheric fluctu- ating reflection channel: Physically based software simulator of the transionospheric channel, Radio Sci- ence, 40 (1), RS1003, doi:10.1029/2004RS003097. GHERM, V.E., N.N. ZERNOV and H.J. STRANGEWAYS (2007): Modelling of the effect of strong scintillation on GPS Signals due to plasma bubbles in the low-latitude iono- sphere, presented at COST 296/IRI meeting (Prague, Czech Republic, 10-14 July, 2007). KOURIS, S.S., LJ. R. CANDER and K.V. POLIMERIS (2005): On the variability of the slab thickness, ESA/ESTEC, 2nd European Space Weather Week, (14-18 November 2005, Noordwjk, The Netherlands), ESA-ESWW2. KOURIS, S.S., K.V. POLIMERIS and LJ. R. CANDER (2006): Specifications of TEC variability, Adv. Space Res., 37, 983-1004, doi:10.1016/j.asr.2005.01.102. KOURIS, S.S., K.V. POLIMERIS, LJ. R. CANDER and L. CIRAO- LO (2008): Solar and latitude dependence of TEC and slab thickness, Journal Atmos. Solar-Terr. Phys., doi:10.1016/j.jastp.2008.03.009. KOURIS, S.S., K.V. POLIMERIS, L. CIRAOLO and D.N. FOTIADIS (2009): Seasonal dependence of TEC and slab thickness, Adv. Space Res., doi:10.1016/j.asr.2008.10.036. KUTIEV, I., P. MARINOV and S. WATANABE (2006): Model of topside ionosphere scale height based on topside sounder data, Adv. Space Res., 37 (5), 943-950, doi:10.1016/j.asr.2006.06.013. KUTIEV, I. and P. MARINOV (2007): Topside sounder model of scale height and transition height characteristics of the ionosphere, Adv. Space Res., 39, 759-766. KUTIEV, I., P. MARINOV, A. BELEHAKI, B. REINISCH and N. JAKOWSKI (2009): Reconstruction of Topside Density profile by using the topside sounder model profiler and digisonde data, Adv. Space Res., 43 (11), 1683-1687. JANSEN, F., P. BECK and LJ.R. CANDER (2006): Space weath- er impacts on aviation and radio communications, in Proceedings of the International Heliophysical Year, First European General Assembly, (10-13 January 2006, Paris, France), pp. 9. MARINOV, P., I. KUTIEV and S. WATANABE (2004): Empirical model of O+-H+ transition height based on topside sounder data, Adv. Space Res., 34, 2015-2022. MAURITS, S.A., V.E. GHERM, N.N. ZERNOV and H.J. STRANGEWAYS (2008): Modeling of scintillation effects on high-latitude transionospheric paths using ionos- pheric model (UAF EPPIM) for background electron density specifications, Radio Science, 43, RS4001, doi:10.1029/2006RS003539. MIDDLETON, H.R., S.E. PRYSE, L. KERSLEY, G.S. BUST, E.J. FREMOUW, J.A. SECAN and W.F. DENIG (2005): Evi- dence for the tongue of ionisation under northward IMF conditions, J. Geophys. Res., 110 (A7), A07301 10.1029/2004JA010800. MIDDLETON, H.R., S.E. PRYSE, A.G. WOOD and R.L. BALT- HAZOR (2008): The role of the tongue-of-ionisation in the formation of the poleward wall of the main trough in the European post-midnight sector, J. Geophys. Res., 113 (A2), A02306, doi:10.1029/2007JA012631. MUHTAROV, P., I. KUTIEV and LJ. R. CANDER (2002): Geo- magnetically correlated autoregression model for short-term prediction of ionospheric parameters, In- verse Problems, 18, 49-65. PERRONE, L. and G. DE FRANCESCHI (1999). A correlation study between time-weighted magnetic indices and the high latitude ionosphere, Physics and Chemistry of the Earth, 24 (4), 389-392. PERRONE, L., M. PIETRELLA and B. ZOLESI (2007): A Predic- tion model of foF2 over periods of severe geomagnetic activity, Adv. Space Res., 39, 674-680. PIETRELLA, M., L. PERRONE, G. FONTANA, V. ROMANO, A. MALAGNINI, G. TUTONE, B. ZOLESI, LJ.R. CANDER, A. BELEHAKI, I. TSAGOURI, S.S. KOURIS, F. VALLIANATOS, J. MAKRIS and M. ANGLING (2009): Oblique-incidence ionospheric soundings over Central Europe and their application for testing now casting and long term pre- diction models, Adv.Space. Res., 43 (11), 1611-1620. PRYSE, S.E, L. KERSLEY, D. MALAN and G.J. BISHOP (2006a): Parameterization of the main ionospheric trough in the European sector, Radio Science, 41, RS5S14, doi:10.1029/2005RS003364. PRYSE, S.E, A.G. WOOD, H.R. MIDDLETON, I.W. MCCREA and M. LESTER (2006b): Reconfiguration of polar-cap plasma in the magnetic midnight sector, Ann. Geophys- icae, 24 (8), 2201-2208, (available on line at www.ann- geophys.net/24/2201/2006/). PRYSE, S.E, H.R. MIDDLETON and A.G. WOOD (2007): Llif yr atmosffer drydanol dros begwn y gogledd: arsylwadau tomograffi radio a SuperDARN, Gwerddon, 1 (2), 38- 56, (ISSN 1741-4261), (available on line at www.gwerd- don.org/2_llif_yr_atmosffer_drydanol_- 73.aspx). Vol52,3,2009 20-09-2009 19:06 Pagina 270 271 Near-Earth space plasma modelling and forecasting PRYSE, S.E., E.L. WHITTICK, A.D. AYLWARD, H.R. MIDDLETON, D.S. BROWN, M. LESTER and J.A. SECAN (2009): Model- ling the tongue-of-ionisation using CTIP with Super- DARN electric potential input: verification by radioto- mography, Ann. Geophysicae, 27, 1139-1152, (available on line at www.ann-geophys.net). RAWER, K., S.S. KOURIS and D.N. FOTIADIS (2003): Vari- ability of F2 parameters depending on modip, Adv. Space Res., 31 (3), 537-541. SENALP, E.T. (2007): Cascade Modeling of Nonlinear Sys- tems, Ph-D Thesis, (Electrical and Electronics Eng., Middle East Technical Univ., Ankara, Turkey). SENALP, E.T., E. TULUNAY and Y. TULUNAY (2008): Total Electron Content (TEC) Forecasting by Cascade mod- eling: a possible Alternative to the IRI-2001, Radio Science, 43, RS4016, doi:10.1029/2007RS003719. STANKOV, S.M., P. MARINOV and I. KUTIEV (2007): Compar- ison of NeQuick, PIM, and TSM model results for the plasma scale and transition heights, Adv. Space Res., 39, 767-773, doi:10.1016/j.asr.2006.10.023. STRANGEWAYS, H.J., V.E. GHERM and N.N. ZERNOV (2007): Modeling and mitigation of the effect of scintillations on GPS, paper at 49th International Symposium EL- MAR-2007, (12-14 September 2007, Zadar, Croatia. Conference Proc.) pp. 69-73, available online in IEEE Xplore, doi:10.1109/ELMAR.2007.4418802. TRISKOVA, L., I. GALKIN, V. TRUHLIK and B.W. REINISCH (2007): Application of seamless vertical profiles for use in the topside electron density modeling, Adv. Space Res., 39, 774-778. TULUNAY, E., Y. TULUNAY, E.T. SENALP and LJ.R. CANDER (2004): Forecasting GPS TEC Using the Neural Net- work Technique «A Further Demonstration», Bulgari- an Geophysical Journal, 30 (1-4), 53-61. TULUNAY, E., E.T. SENALP, S.M. RADICELLA and Y. TULUNAY (2006): Forecasting Total Electron Content Maps by Neur- al Network Technique, Radio Science, 41 (4), RS4016. WHEADON, N.S., R. LARSEN, A. BELEHADI, LJ. R. CANDER, B. ZOLESI, J. BREMER, C. JUREN, I. STANISLAWSKA, D. DIALETIS and M. HATZOPOULOSS (2006): Exploitation of the European Digital Upper Atmospheric Server (DIAS) for Radio Systems Applications, in Proceed- ings of the IET Conference on Ionospheric Radio Sys- tems and Techniques, (18-21 July 2006, London, UK). WHITTICK, E.L., S.E. PRYSE, H.R. MIDDLETON, A.R. LOTINGA and A.D. AYLWARD (2009): The CTIP Model with Su- perDARN electric potential patterns as input: first com- parison of modelled ion densities with radio tomography observations, Earth, Moon and Planets, 104 (1), 45-49, doi:0.1007/s11038-008-9264-9, (available on line at www.springerlink.com/content/53w6472172p73808/). WOOD, A.G., S.E. PRYSE, H.R. MIDDLETON and V.S.C. Howells (2008): Multi-instrument observations of nightside plasma patches under conditions of IMF Bz positive, Ann. Geophysicae, 26, 2203-2216, (available on line at http://www.ann-geophys.net/26/2203/2008/). WRENN, G.L. (1987): Time-Weighted accumulations ap(t) and Kp(t), J. Geophys. Res., 92 (10), 125-10, 129. WU, J. and P.J. WILKINSON (1995): Time-weighted indices as predictors of ionospheric behaviour, J. Atmos. Terr. Phys, 57, 1763-1770. YAPICI, T. (2007): Influences of interplanetary magnetic field on the variability of the aerospace media, MS Thesis, Aerospace Eng., (Middle East Technical Univ., Ankara, Turkey, Sep. 2007). ZERNOV, N.N., V.E. GHERM and H.J. STRANGEWAYS (2008): The scintillation propagation model for the effects of scintillation on transionospheric paths traversing mid- scale ionospheric inhomogeneities, COST296 Work- shop «Ionospheric scintillation: Scientific aspects, Space weather, Application and services», (Notting- ham, UK, 20-22 February, 2008). ZERNOV, N.N., V.E. GHERM and H.J. STRANGEWAYS (2009): On the effects of scintillation of low latitude bubbles on transionospheric paths of propagation, Radio Science, 44, RS0A14, doi:10.1029/2008RS004074. ZOLESI, B., G. FONTANA, L. PERRONE, M. PIETRELLA, V. RO- MANO, G. TUTONE, A. BELEHAKI, I. TSAGOURI, S.S. KOURIS, F. VALLIANATOS, J.P. MAKRIS and M.J. AN- GLING (2008): A new campaign for oblique-incidence ionospheric sounding over Europe and its data applica- tion, J. Atmos. Sol. Terr. Phys., 70, 854-865. ZOLESI, B., and LJ.R. CANDER (2008): From COST 238 to COST 296: Four European COST Actions on Ionos- pheric Physics and Radio Propagation, in AIP Confer- ence Proceedings, 974, 39-46. ZOLESI, B., LJ.R. CANDER, A. BELEHAKI and I. TSAGOURI (2008): Effective sunspot number: A tool for ionos- pheric mapping and modelling, XXIX General Assem- bly of URSI, (Chicago, USA, 07-16 August 2008). Vol52,3,2009 20-09-2009 19:06 Pagina 271