Radio Science

Real-time updating of the Simplified Ionospheric Regional Model for operational applications

Authors


Abstract

[1] A method for mapping of ionospheric conditions over Europe, suitable to be used in real time for operational applications, is described in this paper. The method is based on the Simplified Ionospheric Regional Model (SIRM), a regional model of the standard vertical incidence monthly median ionospheric characteristics that has been updated with real-time (automatic scaled) ionospheric observations to produce nowcasting maps over Europe. As substantial fluctuations from a monthly median regional ionospheric description occur on day-to-day basis, the SIRM results oversimplified a number of the ionospheric phenomena of real significance for radio communications applications. Therefore a rapid conversion of real-time data from four European digisondes (Digital Portable Sounders) to the driving parameters of the Simplified Ionospheric Regional Model is introduced as the real-time SIRM updating (SIRMUP). In this approach, values of the ionospheric characteristics from first-guess model parameters at measurement points are combined with real-time measurements. To assess the qualitative improvements achieved with the real-time SIRM update method, observations of foF2 parameter with SIRMUP predictions were compared for various ionospheric conditions. The simulation shows that the SIRMUP prediction results are much improved comparing to SIRM predictions, especially during large-scale ionospheric disturbances, as well as during quiet conditions, while there was a marginal improvement during localized ionospheric disturbances. In general, the results clearly demonstrate that the proposed procedure of updating SIRM with automatic scaling ionospheric parameters from the four European digisondes has the potential to be used in real time for nowcasting the standard ionospheric characteristics over Europe for operational applications.

1. Introduction

[2] Ionospheric models of the standard vertical incidence (VI) ionospheric characteristics oversimplify a number of the ionospheric phenomena of real significance for radio communications applications. Usually in these models, ionospheric parameters such as critical frequency of the ionospheric F2 layer foF2 and propagation factor M(3000)F2 represent monthly median values and vary as a function of geographic location, local time and solar activity. The Simplified Ionospheric Regional Model (SIRM) belongs to this group of ionospheric models and calculates the values of the key VI ionospheric characteristics such as foF2, M(3000)F2, virtual height of the ionospheric F2 layer h'F2, critical frequencies of the ionospheric F1 and E layers foF1 and foE that are used for prediction of operational parameters of HF telecommunication systems in a restricted area [Zolesi et al., 1993, 1996]. The SIRM was developed under the COST (Cooperation in the field of Scientific and Technical Research) Action 238 PRIME (Prediction and Retrospective Ionospheric Modelling over Europe [Bradley, 1995]) and improved and tested under the COST Action 251 IITS (Improved Quality of Service in Ionospheric Telecommunication Systems Planning and Operation [Hanbaba, 1999]). For the purposes of this work, the improved version of SIRM is used [Zolesi et al., 1999], based on the assumption that at constant local time there are no longitude changes of the ionospheric characteristics and that their diurnal and seasonal variations can be well represented by a Fourier expansion with a relative small number of numerical coefficients. The Fourier coefficients are coming from the analysis of the hourly monthly median values of the ionospheric characteristics measured at the midlatitudes stations over European region and collected under the COST Actions. The COST251 testing procedure consists in comparing measurements of all median hourly data available from a given set of ionospheric stations and the predicted values by different models. According to the results from this comparison, the overall root mean square (RMS) error from SIRM was slightly smaller than RMS error for ITU recommended model [International Telecommunication Union-Radiocommunication (ITU-R), 1994; Levy et al., 1998]. This validation test proofs that SIRM performance is satisfactory for median ionospheric condition description in restricted area of midlatitudes.

[3] All the above indicate that the problem of having a realistic representation of the state of the ionosphere over Europe in real-time is still open. On the other hand, the SIRM provides an efficient and user-friendly software programme with a very simple mathematical formulation of the complex ionospheric media and reduced number of numerical coefficients involved. Therefore one could argue that SIRM, being one of the long term monthly median value models adopted by the scientific community, could be upgraded to a model that meets the need of nowcasting ionospheric conditions over Europe [Zolesi et al., 2002]. In this paper, a real-time updating method of SIRM with autoscaled ionospheric parameters observed by four European digisondes currently capable to provide data in real-time mode has been studied. It is shown that the proposed method introduces a realistic nowcasting of the ionosphere over restricted Europe area, especially during strong geomagnetic storms that make the procedure even more relevant for HF propagation prediction. In the following sections, the methodology of SIRM real-time updating is described and then the simulation results are presented and discussed.

2. SIRM Real-Time Updating Method

[4] The SIRM updating method (SIRMUP) is based on the idea that real time values of foF2 at one location can be determined from the SIRM model by using an effective sunspot number, Reff, instead of the 12-month smoothed sunspot number, R12. The method of determining Reff was introduced and detailed described by Houminer et al. [1993]. The main steps of the methodology are summarized here: Reff is chosen to give the best fit between model calculation and actual measurements obtained from a grid of ionosondes located in the mapping area, refer to hereafter as reference stations. To initiate the procedure, a starting value of foF2, calculated by SIRM model, is used. Then, an iteration procedure is applied to adjust the sunspot number used by SIRM until the mean square error between the SIRM calculation and the real observations is minimized. The sunspot number, giving the minimum mean square error, is called the effective sunspot number. The mean square error is calculated as follows [Houminer et al., 1993]:

equation image

where n is the number of reference stations, foF2obsi is the observed foF2 at the reference station i, foF2calci is the calculated value of foF2 at station i by SIRM. Beginning the procedure for sunspot number 10, increasing to 200 with a step of 10, foF2calci is calculated for the different values of sunspot numbers to determine the minimum mean square error and thus Reff.

[5] To test the reliability of the proposed methodology, the model predicted foF2 values were compared with the observed values resulted from the manual validation of the ionograms recorded at some stations of the mapping area which are refer to hereafter as test stations. In the location of each test station, foF2 was calculated by the SIRM model, using first the observed sunspot number R12, and then the calculated Reff. The relative errors [Houminer et al., 1993] between the observed (manually validated) and the calculated foF2 values were determined using the following equations:

equation image
equation image

where foF2obs is the observed foF2 at the test station, foF2SIRM is the SIRM calculated foF2 at the test station using the observed R12 sunspot number, and foF2SIRMUP is the SIRMUP calculated foF2 at the test station using Reff.

[6] According to the above description, when the criterion

equation image

is valid, the method of real-time updating is successful in the sense that the resulting map in the specific area is more representative of the real ionosphere than the corresponding map resulted from the use of monthly median values.

3. Simulation Results

[7] For operational applications it is important to verify that SIRMUP performance is more successful than SIRM under all possible ionospheric conditions, including quiet intervals, large-scale disturbances such as storm negative phases and daytime positive effects of long duration as well as small-scale effects such as travelling atmospheric disturbances (TADs) and nighttime positive effects. Therefore, observations of the foF2 parameter were compared with SIRMUP predictions for three indicative time intervals and the simulation results are presented and discussed in this section. The coordinates of the reference and test stations involved in this study are given in Table 1, while the geomagnetic conditions during the three test periods as well as the ionospheric conditions observed in the European area are summarized in Table 2. The test stations, from which it was possible to have manual validated ionogram data for each time interval, are also noted in Table 2.

Table 1. Coordinates of the Reference and Test Stations Involved in This Study
 Geographic Latitude, deg NGeographic Longitude, deg E
Reference Stations
Athens38.123.9
Rome41.812.5
Chilton51.6358.7
Juliusruh54.613.4
 
Test Stations
Moscow55.537.3
Sofia42.723.4
Ebro40.50.5
Table 2. List of Simulated Events
Time PeriodGeomagnetic ActivityIonospheric ConditionsTest Stations
24–27 Nov. 2001Storm PeriodModerate disturbancesSofia, Moscow
16–24 Aug. 2001Storm PeriodStrong disturbances with daytime positive effects and prolonged negative effectsSofia
6–10 Dec. 2001Quiet PeriodLarge-scale positive daytime effectsSofia, Ebro

[8] The simulation results for the November 2001 period for two test stations Sofia and Moscow are presented in Figures 1a and 1b, respectively. The time plot of the foF2 (manually validated) observed at the test stations (black line) together with the predicted foF2 by SIRM (green line) and by the real-time updated SIRM (SIRMUP, red line) is presented in the upper panel of these two figures. The monthly median values of foF2 are also given (blue line) to be used as reference level, and be possible therefore to have an estimate of the real ionospheric disturbances recorded in each station. The foF2 deviations from the median values in percentages, δfoF2%, are plotted in the second panel together with the time variation of the quantity (Reff − R12) (green line). The time evolution of the Dst index is over plotted (red line) to give the overall picture of the disturbance level of the magnetosphere-ionosphere system. Finally in the bottom panel, the difference (e1 − e2) is shown. It is expected that the real-time updating is successful when the difference (e1 − e2) posses positive values. During this period an intense geomagnetic storm occurred with the Dst index reaching a minimum of −200nT. Although the intensity of the storm was significant, the ionosphere in the European region was not considerably affected because the onset occurred at 0600UT. At that time the European region was already in the morning sector and it was not affected by the neutral composition disturbance zone generated around midnight and corotated eastward following the Earth's rotation [Prölss, 1993]. Also, as expected from Prölss [1993] phenomenological scenario, as the European region approached local noon, TADs were observed in both test stations. Overall, the ionospheric disturbances represented by δfoF2% were small in magnitude. From the inspection of the variation of δfoF2% and (Reff − R12) with time presented in the second panel of Figures 1a and 1b, it is obvious that there is close relation between the two quantities, indicating a physical relation between the disturbance in ionisation and the deviation of Reff from R12. This is a very interesting feature that will be analysed later, after the presentation of all available results.

Figure 1.

Simulation results for the storm period 24–27 November 2001 storm period at two test stations (a) Sofia and (b) Moscow.

[9] The performance of SIRMUP is strongly depended on the value of Reff. In most of the cases, SIRMUP performance is successful at times when Reff differs considerably from R12. Regarding the relative performance of the method at different sites, one can argue that SIRMUP is more successful at Sofia rather than Moscow station, since the quantity (e1 − e2) appears mostly positive during this time interval at Sofia, but presents several strong negative turnings at Moscow (see the bottom panels of Figures 1a and 1b).

[10] The second storm period used to simulate SIRMUP performance occurred from 16 to 24 August 2001. The results are presented in Figure 2 for Sofia station, in the same format used for Figure 1. The storm started with an initial compression phase in Dst observed on 17 August 2001 at 1200UT. The European region was already in the noon sector and therefore was affected by TADs, causing the strong daytime positive effect, which was observed from Sofia during local afternoon hours. At 2200UT, Dst reached its minimum and consequently the recovery phase started, although considerable substorm activity was recorded by the AE-index (not shown) until midnight. It is possible that this substorm activity caused the generation of a new composition disturbance zone [Tsagouri et al., 2000], producing the strong negative phase the first hours of the 18 August. Consequently, the adjacent TADs generated in the daylight sector probably caused the noticeable positive effect in the morning sector. As the geomagnetic activity recovered the next days, there was a remaining ionospheric disturbance recorded in Sofia of small magnitude. An intense substorm activity recorded by the AE-index from the morning of 21 August until the noon of 23 August produced TADs on the 21 and 22 August in the daylight sector and a noticeable negative phase on the 23 August. The feature that has to be emphasized during this time interval, is an excellent agreement between the δfoF2% and the (Reff − R12) showing that this quantity is indicative of the ionospheric activity. It should be also noticed that most of the time the quantity (e1 − e2) was positive and the performance of SIRMUP was particular successful during periods of large scale ionospheric disturbances causing the negative effects recorded on the 18 and 23 August 2001, as well as of small scale ionospheric disturbances generated by TADs on 17 August 2001. Nevertheless the model failed to predict the positive effect of short duration observed the next day. During periods of successful SIRMUP performance, Reff was much differed than R12.

Figure 2.

Simulation results for the storm period 16–24 August 2001 at the test station of Sofia.

[11] The third period used to test the performance of the SIRMUP model was the interval from 6 to 10 December 2001. The simulation results are presented in Figures 3a and 3b for two test stations, Sofia and Ebro respectively in the same format as Figure 1. During this period a different type of ionospheric activity observed, caused probably by changes in the global wind circulation [Prölss, 1995], which have as result the considerable increase of the foF2 value during day hours, for the first three days of the interval. Although the Dst index did not present any disturbance, the source of the observed ionospheric disturbance is the continuous substorm activity recorded by the AE index. The differences between the relative errors (e1 − e2) plotted in the bottom panel of Figures 3a and 3b have in general smaller values in comparison to the two storm periods presented previously. But it is important to note that the difference (e1 − e2) is positive for most of the time in both stations and that the performance seems more successful for Sofia station than for Ebro. Concerning the later indication, an interesting point is revealed from the SIRMUP curve morphology for Ebro, as it is appeared in the top panel of Figure 3b: during this time interval and especially in the first two days, the SIRMUP values indicate daytime positive effects of short duration, typical of the TADs influence. On the other hand, SIRMUP performance is particular successful during large-scale positive effects observed the first three days of this interval at Sofia and this is probably because Sofia is much closer to the centre of the measurement locations. During nighttime positive effects the performance is not very successful but it is still better than SIRM.

Figure 3.

Simulation results of the quiet period 6–10 December 2001 at two test stations (a) Sofia and (b) Ebro.

[12] To obtain a quantitative estimate of the SIRMUP performance, the scatterplots of e2 versus e1 are shown in Figure 4, for all time periods for which the simulation results were presented. The best fit line is over plotted in each diagram. In all cases the criterion e2 < e1 is statistically satisfied. The correlation coefficient is always less than the unit, getting its larger values (0.60 and 0.75) for the period of December 2001 Moreover, the correlation coefficient for Sofia is always less than its corresponding value for the other two test stations, Moscow and Ebro showing that the SIRMUP performance is more successful at Sofia site than the other two stations. The most probable explanation is that Sofia is located at the center of the mapping area while Moscow and Ebro are at the edges. During the first two intervals, when ionospheric activity was enhanced the SIRMUP performance was much better than SIRM since the correlation coefficient between e1 and e2 is very small. During the third interval, the SIRMUP performance is improved comparing to SIRM, since the correlation coefficient still less than unit, but the overall performance is worse than in previous two cases.

Figure 4.

The scatterplots of e2 versus e1 for the time periods analysed in this work. The best fit line and its equation are also shown.

[13] A very important feature that determined the performance of the real-time updating of SIRM (SIRMUP) is the value used for the Reff. From its definition the Reff is the sunspot number for which the deviation of the observed from the SIRM foF2 values is minimum (equation (1)). It is therefore expected that during periods of relative ionospheric quietness, when the observed foF2 value is much closer to the model predicted median values, the Reff does not differ much from the R12. On the contrary, during periods of ionospheric activity, the Reff should have large difference form R12. Thus, (Reff − R12) should be considered as a regional index of the ionospheric activity for the European region. To further establish this idea, the distribution of the (Reff − R12) parameter for positive and negative cases of the (e1 − e2) is presented in Figure 5a. It is statistically verified that the performance of SIRMUP is better than SIRM since the number of cases for which (e1 − e2) > 0 are always greater than the number of cases for which (e1 − e2) < 0, for all values of (Reff − R12). There is also an indication that the relative performance of the two models depends on the level of ionospheric disturbance as this is expressed by (Reff − R12). To further investigate this point, the distribution of (Reff − R12) is examined only for cases for which (e1 − e2) in absolute values are less than 0.05 (Figure 5b). This will eliminate from the statistical sample cases for which the performance of the two models differs marginally and therefore the results of the distribution should be more important statistically. From Figure 5b it is resulted that SIRMUP performance is strongly improved for cases for which the ionospheric activity is moderated to intense, although during quiet intervals SIRMUP performance is still better than SIRM. In cases of very intense geomagnetic activity, SIRMUP performance is still improved but this result is not statistically important.

Figure 5.

The distribution of the (Reff − R12) quantity for the cases: (a) (e1 − e2) > 0, marked with red and (e1 − e2) < 0, marked with blue, and (b) (e1 − e2) > 0.05, marked with red and (e1 − e2) < −0.05 marked with blue.

[14] The final output from this method are maps of foF2 covering the European area from −5°W to 40°E in longitude and 34°N to 60°N in latitude. A comparison between the ionospheric maps of the foF2 parameter using SIRM and the updated SIRM (SIRMUP) is presented in Figure 6 on 6 December 2001 at 0900UT. During this time the ionosphere over Europe was affected by large-scale effect caused by changes in the global wind circulation. SIRMUP performance according to the e2 < e1 criterion was particularly successful. Indeed the foF2 values in the SIRMUP generated map are greater than those appeared in SIRM map. Also the topology of the ionosphere is much better determined, especially in the southeast region, where a latitudinal dependence of foF2 is well described by SIRMUP.

Figure 6.

The ionospheric map of the foF2 parameter contours in MHz over Europe from −5°W to 40°E in longitude and from 34°N to 60°N in latitude on 6 December 2001 at 0900UT, computed (a) with SIRM and (b) SIRMUP.

[15] A future test of the benefits of SIRM updates can be provided by the comparison with the map derived from measured values (Figure 7) regularly available at http://ionosphere.rcru.rl.ac.uk STIF archive with a time delay of 12 to 24 hours [Cander, 2003]. There are significant similarities in the appearance of the SIRMUP results with the foF2 observation, despite the large difference in the covered areas defined by number of real-time operating European digisonds used for model updating in case of SIRMUP and by number of ionospheric stations sending rapidly past measured data for Short-Term Ionospheric Forecasting in case of the STIF.

Figure 7.

The ionospheric map over Europe from −10°W to 90°E in longitude and from 30°N to 70°N in latitude derived from measured values of foF2 for 6 December 2001 at 0900UT. The red crosses indicate measured values at ionospheric stations.

4. Discussion and Conclusions

[16] The aim of this work was to develop a technique for the real-time mapping of the ionosphere over Europe with sufficient accuracy that could be used for operational nowcasting applications.In Europe there are four digisondes operated regularly in Athens, Rome, Juliusruh and Chilton. These stations are Digisonde Portable Sounders (DPS4) and provide the ionospheric parameters in real-time, using the ARTIST software for the automatic scaling of the ionograms [Reinisch and Huang, 1983]. The geographical distribution of these stations is satisfactory covering most part of the European midlatitude ionosphere considered relevant for the SIRMUP methodology presentation in this paper.

[17] On the other hand, a group of ionospheric models providing the long term monthly median values of the ionospheric parameters, using as input the local time, latitude and solar activity are currently available. All these models show more or less the same performance [Levy et al., 1998], giving satisfactory results during quiet times, but deviate considerably from the observed values during disturbed ionospheric conditions. SIRM belongs to this group of ionospheric models. The objective of this study was to show that a method based on SIRM, updated using automatic scaling values of foF2 from the four European digisondes, can describe successfully the state of the ionosphere over Europe during quiet and disturbed conditions and on the other hand can be operated in real-time, covering the corresponding need on this field in Europe.

[18] The results presented show a significant improvement of SIRMUP performance comparing to SIRM, during quiet intervals and also during large-scale ionospheric disturbances. A marginal improvement during localized ionospheric disturbances is also reported. The latter could be explained by recalling the proposed methodology. SIRMUP is very much dependent on the calculated value of Reff. By its definition Reff is a regional index of ionospheric activity and its derivation is based on the value of the automatic scaled foF2 in the four reference stations. Ionospheric disturbances caused by large-scale phenomena are reflected in Reff and consequently in SIRMUP predictions. On the other hand, small-scale effects are not imposed in Reff unless the source of the disturbance is located near to the reference station and vice versa: small scale effects recorded in Reff may lead to deviations from the observed values in some test points. This might be a possible explanation for the weakness of the SIRMUP nowcasting during daytime positive effects of short duration or nighttime positive effects. Moreover, the SIRMUP performance was always more successful for the test stations located near the center of the mapping area than for those located at the edges, indicating that the distribution of the reference stations is a significant and fundamental factor for the success of the model. This is in a good agreement with the Gibson and Bradley [1991] findings on the relationship between numbers and positions of measurement locations needed for the generation of instantaneous maps.

[19] Another limitation in the accuracy of SIRMUP predictions comes from the performance of ARTIST software under certain conditions, such as interference in the ionogram trace. In such cases, ARTIST fails to scale correctly the ionogram, leading to an error in foF2 and consequently in Reff. However, since the Reff is calculated from automatic scaled ionograms from the four reference stations the contribution of individual errors in the total result is expected to be small.

[20] Future improvements on the proposed methodology could be possible if the calculation of Reff could be based on a more dense network of digisondes, since the number of reference stations and its distribution in the mapping area affects the accuracy of real-time updating method, a remark which is in confirmation with Houminer et al. [1993]. Additional vertical-incidence ionosonde DPS4 deployment at Pruhonice (49.98°N, 14.55°E) in foreseeable future will be significant step in SIRMUP optimum mapping application.

Acknowledgments

[21] We would like to thank Juergen Bremer and David Altadill for providing us with sounding files from Juliusruh and Ebro Digisondes. This work is a part of the European Project COST271 and was partly supported by the ESA Space Weather Pilot Project GIFINT.

Ancillary