Annals 48, 2, 2005defdef 321 ANNALS OF GEOPHYSICS, VOL. 48, N. 2, April 2005 Key words electron density model – vertical inci- dence ionograms – ionospheric data ingestion 1. Introduction One of the most unpredictable sources of er- ror for single frequency satellite navigation and positioning systems is due to the ionosphere. A good knowledge of the behavior of the ionos- pheric regions is essential to obtain a better rep- resentation of parameters like the Total Electron Content (TEC) which is directly proportional to the time delay of electromagnetic signals. A technique to reconstruct the spatial and temporal structure of the electron concentration in the ionosphere has been developed using the NeQuick ionospheric electron density model driven by an «effective ionization parameter» called Az (Nava et al., 2003). NeQuick is a quick-run model for ionospheric applications developed at The Abdus Salam ICTP in Trieste (Italy) and the University of Graz (Austria). It has been used by the European Space Agency - ESA satellite navigation and positioning pro- grams and adopted by Recommendation P.531- 6 of the ITU-R (International Telecommunica- tion Union, Radiocommunication sector) (now superseded by P.531-7; ITU, 2001). Az values are determined minimizing the dif- ferences between experimental and modeled ver- tical TEC global maps. The result is a global grid of Az values for a specific condition which al- lows us to calculate with the NeQuick model the electron density at any point in the ionosphere including the F2 peak parameters values. The aim of this paper is to compare these cal- culated F2 peak values with independent and si- multaneous global observations in order to vali- date the technique. These measurements, the criti- Validation of a method for ionospheric electron density reconstruction by means of vertical incidence data during quiet and storm periods Gloria Miró Amarante, Sandro M. Radicella, Bruno Nava and Pierdavide Coïsson The Abdus Salam International Centre for Theoretical Physics (ICTP), Trieste (Italy) Abstract A preliminary validation of the technique developed using the NeQuick ionospheric model and the «effective ion- ization parameter» Az, based on vertical total electron content data ingestion, was carried out in a previous study. The current study was performed to extend the analyzed conditions and confirm the results. The method to validate this technique is based on a comparison between hourly F2 peak values measured with Vertical Incidence (VI) soundings and those calculated with the new technique. Data corresponding to different hours and seasons (equinox, summer solstice, and winter solstice) during the period 2000-2003 (high and medium solar activity conditions) were compared for all available ionosonde stations. The results show a good agreement between foF2 and hmF2 values obtained with the new technique and measurements from vertical incidence soundings during quiet and storms conditions. Mailing address: Dr. Gloria Miró Amarante, The Ab- dus Salam International Centre for Theoretical Physics (ICTP), Strada Costiera 11, 34014 Trieste (Italy); e-mail: amarante@ictp.it 322 Gloria Miró Amarante, Sandro M. Radicella, Bruno Nava and Pierdavide Coïsson cal frequency in the F2 region (foF2) and the height of the maximum of ionization (hmF2), are ob- tained experimentally from ground-based vertical ionospheric sounders located around the world. A preliminary validation was done (Miró Amarante et al., 2003) using data correspon- ding to four quiet days in the year 2002 (high solar activity conditions). The main results and conclusions of this analysis will be included in this paper. 2. Ionospheric parameters measured with vertical incidence soundings The DIDbase software developed by Lowell University lets us extract ionograms from a net- work of digisondes distributed around the world. This possibility allows manual scaling of the available ionograms to measure the real val- ues of the main ionospheric parameters (critical frequency, height, Ionospheric Total Electron Content - ITEC: Huang and Reinisch, 2001). The analysis has been done for different conditions summarized in table I. For each pe- riod, all the available ionospheric stations were selected (fig. 1). The number of locations varies from 13 during the period of March 2000 up to 27 for the 3rd of June 2002. The manual scaling of the ionograms (al- most 8300 ionograms) was done for each con- sidered case and foF2 and hmF2 were extracted to be compared with the reconstructed values. 3. Ionospheric parameters calculated with the reconstruction technique Two global grids corresponding to foF2 and hmF2 were obtained from the NeQuick model driven by the «effective ionization parameter» Az. This parameter is calculated by means of the technique described in Nava et al. (2003) in such a way that the differences between experimental and NeQuick model vertical TEC are minimum. The experimental vertical TEC maps used for the minimization are generated with two different techniques. The first is the one developed by CODE (Centre for Orbits DEterminations, http://www.cx.unibe.ch/aiub/ionosphere.html) with two hour intervals and the second one was generated by the University of La Plata, Argenti- na with one hour intervals (Meza et al., 2002; and Brunini et al., 2004). The calculated F2 peak values at the avail- able digisonde locations (fig. 1) are extracted Fig. 1. Observing sites available for the periods ana- lyzed. Table I. List of selected periods for the validation of the reconstruction method. Day Month Year Ap Dst Sunspot number 25 03 2000 8 − 9 185 26 03 2000 3 − 1 170 27 03 2000 4 1 155 28 03 2000 4 1 169 15 04 2002 6 − 8 138 03 06 2002 10 − 21 133 11 10 2002 6 − 38 121 25 11 2002 15 − 46 56 19 11 2003 12 − 179 70 20 11 2003 150 − 152 90 21 11 2003 42 − 76 97 22 11 2003 30 − 63 83 23 11 2003 22 − 38 109 323 Validation of a method for ionospheric electron density reconstruction by means of vertical incidence data from these two global grids by interpolation since the Az maps have a grid spacing of 2.5° in latitude and 5° in longitude. 4. Comparison between measured and calculated peak values As an example, the critical frequency foF2 and the maximum height hmF2 for Millstone Hill (middle latitude Northern Hemisphere) and Gra- hamstown (middle latitude Southern Hemi- sphere) are shown in fig. 2a-d (quiet period) and fig. 3a-d (storm period). Calculated F2 peak val- ues using CODE (foF2CODE) and La Plata (foF2LAPLATA) vertical TEC maps are plotted together with the experimental ones (foF2SAO, obtained with Sao-Explorer, http://ulcar.uml.edu/ SAO-X/SAO-X.html). The figures show a good agreement between both foF2 data sets (foF2CODE-foF2SAO and foF2LAPLATA-foF2SAO) during quiet condi- tions while the differences between calculated and experimental values increase for geomagnet- ic disturbed conditions. The results are consider- ably worse in the case of hmF2. However, this is not surprising. NeQuick uses the CCIR (ITU-R) maps for foF2 and M(3000)F2 and an internal map for foE to calculate hmF2 by means of Du- deney’s form of the Bradley and Dudeney (1973) formula (see Dudeney, 1983) which works quite well for average (monthly median) conditions but can give larger errors in individual cases. Minimizing differences in electron content also means some minimization for foF2 but hmF2 is not affected. The height error of true height analysis of ionograms can also be quite large. The left panel of fig. 4 presents the scatter plot of foF2 (F2 peak electron density) recon- structed with the NeQuick model against the corresponding foF2 ionosonde measurements for quiet periods (2000 and 2002). This example Fig. 2a-d. foF2 (a) and hmF2 (c) for Millstone Hill (middle latitude Northern Hemisphere); foF2 (b) and hmF2 (d) for Grahamstown (middle latitude Southern Hemisphere), during March 2000 (quiet period). b c d a Fig. 4. Linear regression fitting for reconstructed and experimental foF2 during quiet (left) and storm (right) con- ditions. 324 Gloria Miró Amarante, Sandro M. Radicella, Bruno Nava and Pierdavide Coïsson Fig. 3a-d. foF2 (a) and hmF2 (c) for Millstone Hill (middle latitude Northern Hemisphere); foF2 (b) and hmF2 (d) for Grahamstown (middle latitude Southern Hemisphere), during November 2003 (storm period). corresponds to CODE experimental vertical TEC maps. The scatter plot shows a high degree of correlation between the two independent es- timates of foF2. The line drawn corresponds to the best-fit line. Examination of the intercept of the best-fit line shows that on average the foF2NeQuick values exceed those from digison- des by ∼1.16 MHz during quiet period. The a c d b 325 Validation of a method for ionospheric electron density reconstruction by means of vertical incidence data storm conditions are also shown on the right panel in fig. 4. This example corresponds to 20th November 2003 (Ap index=150) and shows how the agreement between both data sets is also ob- tained for a very disturbed day. It is important to point out that this storm occurred under middle solar activity conditions (sunspot number = 90) and the analysis was done considering together all the ionospheric effects (positive and negative storms) distributed around the world. The selected data are classified to create dis- tributions with different conditions of solar ac- tivity, vertical TEC mapping technique and qui- et or storm periods. The results of the linear re- gression fitting corresponding to each distribu- tion are summarized in tables II and III for foF2 and hmF2 respectively. The correlation coefficients R make clear the difference in accuracy for reconstruction of foF2 and hmF2. In the case of the critical frequency the correlation coefficient is approximately 0.90 which means that the fitting between both data set is excellent. However this value does not ex- ceed 0.70 for the peak height possibly for the reasons indicated above when discussing the re- sults shown in figs. 2a-d and 3a-d. The foF2 re- construction is also good during the studied storm period with values higher than 0.86. It must be noted that regardless of the actual val- ues of the correlation coefficient all the results indicate a very high statistical significance. The comparison between the two vertical TEC mapping techniques (CODE and La Plata) does not show a clear dependence on the tech- Table III. Linear regression fitting results for hmF2 (hmF2NeQuick = A∗hmF2Ionosonde + B; R is the correlation coefficient and N the number of selected data). hmF2 (km) La Plata CODE Period A B R N A B R N Calm 2000 0.34 187.7 0.53 1094 0.35 183.9 0.48 546 2002 No data 0.55 128.6 0.60 529 19.11.2003 0.75 70.0 0.70 325 0.72 72.5 0.68 165 20.11.2003 0.62 129.4 0.50 279 0.56 141.9 0.46 140 Storm period 21.11.2003 0.72 90.1 0.50 242 0.80 68.2 0.52 113 22.11.2003 0.76 80.1 0.66 324 0.72 83.1 0.67 164 23.11.2003 0.72 90.2 0.68 323 0.62 113.3 0.66 156 Table II. Linear regression fitting results for foF2 ( foF2NeQuick = A∗ foF2Ionosonde + B; R is the correlation co- efficient and N the number of selected data). foF2 (MHz) La Plata CODE Period A B R N A B R N Calm 2000 0.98 0.20 0.89 1096 1.01 − 0.10 0.91 547 2002 No data 0.94 1.44 0.91 936 19.11.2003 0.90 − 0.06 0.91 469 0.96 − 0.04 0.93 241 20.11.2003 0.96 0.12 0.89 402 1.00 0.24 0.89 204 Storm period 21.11.2003 0.76 0.74 0.86 379 0.79 0.98 0.87 181 22.11.2003 0.91 0.17 0.88 463 0.96 0.23 0.89 232 23.11.2003 0.86 0.48 0.89 457 0.92 0.53 0.90 225 326 Gloria Miró Amarante, Sandro M. Radicella, Bruno Nava and Pierdavide Coïsson nique of the vertical TEC map used to derive Az values. The number of points corresponding to La Plata technique is higher because these maps are obtained hourly. The dependence on solar activity is only no- ticeable for hmF2 where the fitting improves when solar activity decreases. However, further analysis considering low solar activity and storm conditions under very high solar activity should be done in the future. 5. Conclusions The comparison of F2 peak values generat- ed by the new technique of 3D reconstruction of the electron density and the corresponding vertical incidence ionosonde data covering a wide area of different geographical, seasonal and hourly conditions indicates that: – The critical frequency of the F2 layer shows a very good agreement between both data sets during high and middle solar activity quiet periods. – This agreement is also good for the No- vember 2003 storm considering together all the ionospheric effects (positive and negative storm effects) distributed around the world. – The hmF2 parameter shows a worse agreement than foF2 values and the worst results have been found for high solar activity. – The results obtained are not dependent on the technique of the vertical TEC maps used, CODE and La Plata. Therefore, it has been demonstrated that the new technique can be used to build scenarios that reproduce global ionospheric conditions in a realistic way. Acknowledgements The Aeronomy and Radiopropagation Labo- ratory group of The Abdus Salam International Centre for Theoretical Physics is grateful to the University of La Plata for providing their verti- cal TEC maps and to the ionospheric groups (Lowell University (U.S.A.), National Observa- tory of Athens (Greece), Radio Research Labo- ratory (Sinpil Seolsung Ichon Kyunggi, Repub- lic of Korea), Rhodes University (Grahamstown, South Africa), Communications Research Labo- ratory (Tokyo, Japan), Rutherford Appleton Laboratory (Chilton, United Kingdom), Istituto Nazionale di Geofisica e Vulcanologia (Rome, Italy), INTA (Huelva, Spain), Ebro Observatory (Spain), CASLEO (San Juan, Argentina)) from whom we received data used in this study. One of the authors (G.M.A.) also wants to thank to the Marie Curie Fellowship of the European Community Fifth Framework Programme under contract No. HPMF-CI-2002-01867 for support- ing this research. REFERENCES BRADLEY, P.A. and J.R. DUDENEY (1973): A simple model of the vertical distribution of electron concentration in the ionosphere, J. Atmos. Terr. Phys., 35, 2131-2146. BRUNINI, C., A. MEZA, F. AZPILICUETA, A. VAN ZELE, M. GENDE and A. DIAZ (2004): A new ionosphere moni- toring technology based on GPS, Astrophys. Space Sci., 290 (3), 415-429. DUDENEY, J.R. (1983): The accuracy of simple methods for determining the height of the maximum electron con- centration of the F2-layer from scaled ionospheric characteristics, J. Atmos. Terr. Phys., 45, 629-640. HUANG, X. and B.W. REINISCH (2001): Vertical electron content from ionograms in real time, Radio Sci., 36 (2), 335-342. ITU (2001): Ionospheric propagation data and prediction methods required for the design of satellite services and systems ionosphere characteristics, Recommenda- tion P.531-6, approved in 2001-02, managed by ITU-R Study Group SG3 (version 6 superseded in 2004 by version 7). MEZA, A., C.A. BRUNINI and A. KLAEUSBERG (2002): Glob- al behaviour of the ionosphere electron density using GPS observations, Adv. Space Res., 30 (2), 307-312. MIRÓ AMARANTE, G., S.M. RADICELLA, B. NAVA and P. COÏSSON (2003): Validation of a new method for ionos- pheric electron density reconstruction by means of ver- tical incidence data, in Proceedings of ‘Atmospheric Remote Sensing using Satellite navigation Systems’, Special Symposium of the URSI Joint Working Group, Matera (FG), Italy, 335-342. NAVA, B., P. COÏSSON, G. MIRÓ AMARANTE and S.M. RADI- CELLA (2003): A new model assisted method for iono- sphere electron density reconstruction, in Proceedings of ‘Atmospheric Remote Sensing using Satellite navi- gation Systems’, Special Symposium of the URSI Joint Working Group, Matera (FG), Italy, 343-348. (received October 18, 2004; accepted March 19, 2005)