An attempt to validate HF propagation prediction conditions over Sub Saharan Africa

Transcription

1 SPACE WEATHER, VOL. 9,, doi: /2010sw000643, 2011 An attempt to validate HF propagation prediction conditions over Sub Saharan Africa Mpho Tshisaphungo, 1,2 Lee Anne McKinnell, 1,2 Lindsay Magnus, 3 and John Bosco Habarulema 1,2 Received 17 November 2010; revised 2 May 2011; accepted 17 May 2011; published 4 August [1] Ionospheric behavior plays an important role in high frequency (HF) radio propagation, which subsequently provides an opportunity for studying ionospheric variability and space weather effects. The continuous change in ionospheric conditions caused by space weather strongly affects HF propagation. The use of HF communication is still very relevant over the African continent, as seen by the requirements for services provided by the Regional Warning Center for Space Weather in Africa, and this has necessitated an investigation into the prediction capabilities of the existing HF propagation models currently used over this region. This paper presents the validation of HF propagation conditions through the ionosphere by using the Ionospheric Communication Enhanced Profile Analysis and Circuit (ICEPAC) model with real time data from international beacons located in Africa. The HF propagation results presented are for the circuits Ruaraka (5Z4B), Kenya (1.24 S, E), and Pretoria (ZS6DN), South Africa (25.45 S, E) to Hermanus (ZS1HMO), South Africa (34.27 S, E). The potential of this model as compared to real time data in terms of propagation condition predictions is illustrated. An attempt to draw conclusions for the future improvement of HF propagation models is also presented. Results show that ICEPAC performs better for the 5Z4B ZS1HMO than for the ZS6DN ZS1HMO circuit, although it does, in general, provide a low accuracy prediction compared to the real time data. Thus certain parameters need to be investigated further for future improvement in the performance of the ICEPAC model over Africa. Citation: Tshisaphungo, M., L. A. McKinnell, L. Magnus, and J. B. Habarulema (2011), An attempt to validate HF propagation prediction conditions over Sub Saharan Africa, Space Weather, 9,, doi: /2010sw Introduction [2] Space weather is a major influence of ionospheric variability which in turn affect HF propagation. The variability of the ionospheric region ( km) may cause unstable conditions for HF radio communications. The variations in this region are known to occur on both a short term and long term basis. Short term ionospheric behavior include daily variations while on a relatively long term seasonal and solar cycle trends are evident in ionospheric measurements [e.g., Davies, 1989]. Knowledge of ionospheric behavior is crucial for understanding and predicting HF propagation conditions. Over the African continent HF propagation forms a significant part of space weather studies due to the real requirement for HF communication in developing countries. HF communication is 1 Space Science, South African National Space Agency, Hermanus, South Africa. 2 Department of Physics and Electronics, Rhodes University, Grahamstown, South Africa. 3 South African Square Kilometer Array, Pinelands, South Africa. still a basic need to many users over the African region and it is important to make accurate space weather predictions and forecasts. The South African National Space Agency (SANSA) operates the Regional Warning Center (RWC) for space weather over Africa, and one of the RWCs major clients requires an HF prediction service which we are currently providing. Through providing this service we have determined that the topic covered in this paper is a major concern to African space weather users and an investigation into the reliability and accuracy of the current HF propagation tools is of paramount importance to Space Weather users in Africa. There are existing prediction programs for HF propagation conditions, but to date none have been fully developed specifically for the African region. This paper focuses on the validation of the Ionospheric Communication Enhanced Profile Analysis and Circuit (ICEPAC) model over selected African HF propagation circuits. The prediction accuracy of ICEPAC in the Southern hemisphere may not necessarily be the same as for the Northern hemisphere where the model was developed, since it is known that the model had little data for the high latitude and Southern hemisphere Copyright 2011 by the American Geophysical Union 1of7

2 Figure 1. A worldwide map showing the beacon transmitters locations indicated by call signs, 5Z4B and ZS6DN, with a receiver of call sign ZS1HMO. regions available for its development [Lane, 2005b]. It should be noted that currently ICEPAC is the propagation prediction program that is used extensively over Africa, and therefore, while ICEPAC may not be the software of choice it is necessary to demonstrate the need for an improved prediction program for what is currently used over Africa. [3] Recently, a study was undertaken to assess the performance of the ICEPAC model using the maximum usable frequency (MUF) propagation parameter over Central Europe [Zolesi et al., 2008; Pietrella et al., 2009]. The oblique ionospheric radio sounding measurements were considered for the circuits Inskip, UK (53.5 N, 2.5 W) and Rome, Italy (41.8 N, 12.5 E) and Inskip and Chania, Greece (35.7 N, 24.0 E) for periods which fell within the declining phase of solar cycle 23 (2003 and 2004). [4] In this paper a similar study is performed for two propagation circuits within the African region. The transmitting stations used are Ruaraka (5Z4B), Kenya (1.24 S, E) and Pretoria (ZS6DN), South Africa (25.45 S, E) while the station at Hermanus (ZS1HMO), South Africa (34.27 S, E) was used for the receiver. The signal to noise ratio (SNR) obtained from the international beacon project (IBP) measurements is used to assess the propagation performance of the ICEPAC model over the chosen circuits. The predictions were made for the period between September 2008 and June 2009, which is a period of minimum solar activity. This paper presents the first attempt to validate HF propagation models over the African region (Southern Hemisphere) by comparing the SNR data from the ICEPAC model and IBP measurements. The SNR parameter has been specifically used because it is the output measure of propagation conditions from the IBP system. 2. Data Sources 2.1. ICEPAC Model [5] The ICEPAC model is a full performance system for radio communication circuits used to predict HF propagation conditions. It is an improved model of the Ionospheric Communications Analysis and Prediction Program (IONCAP) after including the ICED (ionospheric conductivity and electron density) profile model. The ICED included a better representation of the auroral trough (F. G. Stewart, Ionospheric Communications Enhanced Profile Analysis and Circuit (ICEPAC) prediction program, 2008; available at html). The ICEPAC model is usually classified as a longterm prediction model [Zolesi et al., 2008]. This model was developed using worldwide data but had little data for the high latitude regions and for the Southern Hemisphere [Lane, 2005a]. In this paper, two indices: sunspot number (SSN) and effective geomagnetic activity index (Qfe), were used as inputs to generate the predictions. There are several output parameters obtained from this model such as the radiation angle, SNR, MUF, transmitter and receiver gains etc. Different output parameters can be used to validate HF propagation models depending on available measured data. For example, Stocker et al. [2003] and Walden [2010] used parameters like noise power, lower and upper decile of SNR etc. to make a comparison between measurements and predictions by the Voice of America Coverage Analysis Program (VOACAP). The SNR output parameter is used in this paper to investigate the ability of ICEPAC to predict HF propagation conditions The International Beacon Project [6] The Northern California DX Foundation (NCDXF) and the International Amateur Radio Union (IARU), constructed and operate a worldwide network of HF radio beacons on 14.10, 18.11, 21.15, 24.93, and MHz. These beacons are important for both Amateur and commercial HF radio users to assess the current condition of the ionosphere. The transmitting power used is 100 watts. These beacons are located worldwide as indicated by their call signs in Figure 1. [7] This paper focuses only on two transmitting stations: 5Z4B (Ruaraka, Kenya) and ZS6DN (Pretoria, South Africa) and the receiver station which is ZS1HMO (Hermanus, 2of7

3 Figure 2. Measured and predicted SNR for circuit 5Z4B ZS1HMO for December South Africa). These are the only international beacon stations located in Africa which limited this study to two propagation circuits. The receiver is an ICOM IC 728 with MFJ G5RV multiband antenna located 12 meters above the ground which receives signals in all five frequency bands. It was installed in an inverted vee configuration with an optimum operating frequency of 14.1 MHz [Mudzingwa, 2009]. The transmitter (5Z4B) is a Cushcraft R5 vertical set which is 10 meters above the ground. It has about a 360 area coverage with relatively low elevation angle. The second transmitter (ZS6DN) is a multiband vertical antenna about 6 meters high. We recognize that ICEPAC is not the best model for generating the SNR parameter due to noise associated with predictions and measurements, however, it was not possible to obtain MUF values from our beacon measurements due to the fixed frequency observations of the beacon, and therefore, SNR values were used for the analysis presented. Also, all results presented in this paper were obtained at the frequency of 14.1 MHz, due to this being the frequency for which the largest data set of recorded measurements was available. 3. Comparison of ICEPAC Predictions With IBP Measurements [8] The data used comprises two months of each of 2008 (September and December) and 2009 (March and June). Given that ICEPAC is a long term prediction model, four months data analysis may be a short term period to give a conclusive prediction ability of the model [e.g., Walden, 2010; McNamara et al., 2006]. However, the data available does provide an insight into the short term analysis of the measured and predicted data. These particular months were selected to take into account the solstice and equinox analysis of HF propagation conditions during these seasons. Determining the performance of models with respect to real data during seasons of high and low ionospheric variability is normally done by considering representations of equinox and solstice periods [e.g., Habarulema et al., 2009]. It should be noted that measured data from the transmitting stations is archived at the Hermanus monitoring station for the IBP transmitting stations Daily HF Propagation Conditions [9] The results shown in Figures 2 and 3 illustrate the daily variation of the HF propagation conditions from 23 to 30 December 2008 for circuits Ruaraka (5Z4B), Kenya to Hermanus (ZS1HMO), South Africa and Pretoria (ZS6DN), South Africa to Hermanus, South Africa. In Figure 2 the measured data is compared with the predicted data to assess the performance of the ICEPAC model in predicting HF propagation for the 5Z4B ZS1HMO circuit. The results show an improved performance of the ICEPAC model during most hours of the day, while there appears to be a degradation in performance during the early hours of the day, from about 04:00 to 08:00 UT. However, on average the predicted data seems to be closely related to the measured data. [10] Figure 3 shows the daily variation of the HF propagation conditions for the ZS6DN ZS1HMO circuit. There is a similar trend observed between measured and predicted data during the midday hours from about 07:00 to 3of7

4 Figure 3. Measured and predicted SNR for circuit ZS6DN ZS1HMO for December :00 UT. However, the most noticeable phenomenon in Figure 3 is the sudden decline and increase in the data forming sharp peaks at 12:00 14:00 UT for all the days considered. These peaks are more significant in the predicted data than the measured data. The probable cause of the peaks may be due to the occurrence of skip fading. According to McNamara [1991], skip fading occurs when the operating frequency is exactly equal to the MUF for a given propagation circuit. The predicted MUF presented in Figure 4 for the similar period when the peaks occurred (as seen in Figure 3) indicate an agreement with this theory to some degree. Looking at Figure 4 at 12:00 14:00 UT, which is approximately the time of the peak occurrence, the frequency has a range of MHz where the operating frequency (of 14.1 MHz) used in this study lies. [11] To get a quantitative measure of the statistical accuracy for the ICEPAC model in predicting the propagation conditions for the two circuits considered, the root mean square error (RMSE) method has been used. In this case the RMSE is defined as vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi u 1 X N RMSE ¼ t R N m s R p 2 s ð1þ where N is the number of data points, and R s m and R s p are measured and predicted SNR values, respectively. Various other statistical studies have used RMSE and correlation coefficients to estimate the accuracy of data driven models [e.g., Zolesi et al., 2008; Habarulema et al., 2009]. Table 1 shows the RMSE values and correlation coefficients (corrcoef) between the measured and predicted i¼1 SNR data for December 2008 for the two propagation circuits. Based on the limited presence of measured data in the ICEPAC model from the Southern Hemisphere, the corrcoef of less than 0.5 will be considered as a poor correlation while a good correlation will be considered to be greater than 0.5 [e.g., Lunin and Woolfson, 1993]. Table 1 shows high RMSE values for the ZS6DN ZS1HMO circuit compared to the RMSE for the 5Z4B ZS1HMO circuit. Low RMSE values indicate the closeness Figure 4. Predicted MUF for ZS6DN ZS1HMO circuit for 23 December of7

5 Table 1. RMSE and Correlation Coefficient Values of the Daily Measured and Predicted SNR for Circuits 5Z4B ZS1HMO and ZS6DN ZS1HMO During December 2008 Dates a RMSE (db) Correlation Coefficient 5Z4B ZS1HMO ZS6DN ZS1HMO 5Z4B ZS1HMO ZS6DN ZS1HMO a Date format is day month year. of the predicted and measured data in this case. Low RMSE for the 5Z4B ZS1HMO circuit directly translates into relatively high correlation as seen in Table 1 except on 24 December 2008 where a negative correlation is observed. Considering the period analyzed, average values indicate a corrcoef of 0.43 for the 5Z4B ZS1HMO circuit compared to 0.34 for the ZS6DN ZS1HMO circuit. This shows that for both circuits, the ability of the ICEPAC model to predict HF propagation is less than Monthly Median HF Propagation Conditions [12] Figures 5 and 6 show the monthly median representation of the HF propagation conditions for the 5Z4B ZS1HMO and ZS6DN ZS1HMO circuits, respectively. These results show that ICEPAC does not necessarily follow the variational trend of the measured median SNR values. The predicted SNR in Figure 5 decreases from about 18 db to 5 db on average at 06:00 to 09:00 UT which is not the case for measured SNR data. However, Figures 5a and 5c show a strong signal during the early and late hours of the day and a weak signal during the midday hours for both predicted and measured SNR values. But the difference between the two still remains significant. In Figure 6 there is a strong signal during the midday hours. There is a slight agreement in the variational trend in Figures 6b 6d from 08:00 to 12:00 UT. The computation of RMSE and corrcoef is again used to statistically analyze the monthly median performance of the ICEPAC model. Figure 5. The monthly predictions for the 5Z4B ZS1HMO circuit at 14.1 MHz for four months: September 2008, December 2008, March 2009, and June of7

6 Figure 6. The monthly predictions for the ZS6DN ZS1HMO circuit at 14.1 MHz for four months: September 2008, December 2008, March 2009, and June Table 2 shows the computed RMSE and corrcoef for the two paths. The RMSE for the ZS6DN ZS1HMO circuit is relatively large compared to the 5Z4B ZS1HMO circuit. Both daily and median HF propagation results show an agreement where ICEPAC performs reasonably for the long propagation circuit (5Z4B ZS1HMO) compared to the short circuit (ZS6DN ZS1HMO). 4. Conclusion [13] The aim of this paper was to demonstrate the accuracy in using ICEPAC to predict HF propagation conditions over Africa. This has been achieved by utilizing the limited data available from single frequency beacons. The HF propagation conditions for two paths in Sub Saharan Africa have been analyzed. This was performed by comparing the SNR values from the IBP measurements with predictions generated by the ICEPAC model. The agreement between measurements and predictions are much better for daily predictions as compared to monthly predictions, although the model still has a low accuracy. Two stations over the whole of Sub Saharan Africa are not enough to provide a good understanding of the model performance. Therefore, there is a need for large area data coverage over the African region to conclusively understand the ability of the ICEPAC model to predict HF propagation conditions for Africa. To achieve this, there is a proposed beacon project for Africa that will assist to collect the required data. For this preliminary study, we have used data for only four months which may not be enough for validating a long term prediction model [e.g., Zolesi et al., 2008]. Long term data for both low and high solar activity periods is important for analyzing HF propagation models. Since the model included little data over the African region during the development, a suggestion may be to look at ways of improving the model by including new data. However, this is difficult to achieve since there is no long term existing ionospheric database over the African sector. Our long term objective is to validate existing HF prediction models over the African region with a view to developing a model that effectively represents Africa s ionospheric conditions. We believe that the continuation of this kind of study and the eventual improvement of HF propagation prediction models for the African region will benefit Space Weather users in Africa, and greatly enhance the products offered by the RWC for Africa. Currently a display of the real time propagation data from the IBP can be viewed on our website spaceweather.hmo.ac.za. Table 2. RMSE and Correlation Coefficient Values for the Monthly Median SNR Values Between Two Circuits: 5Z4B ZS1HMO and ZS6DN ZS1HMO Months Correlation RMSE (db) Coefficient 5Z4B ZS6DN 5Z4B ZS6DN September December March June of7

7 [14] Acknowledgments. Great thanks to Kobus Olckers for the assistance with the Qfe index data. References Davies, K. (1989), Ionospheric Radio, Peter Peregrinus, London. Habarulema, J. B., L. A. McKinnell, and B. D. L. Opperman (2009), A recurrent neural network approach to quantitatively studying solar wind effects on TEC derived from GPS: Preliminary results, Ann. Geophys., 27(5), Lane, G. (2005a), Improved guidelines for automatic link establishment operations, paper presented at Ionospheric Effects Symposium, Alexandria, Va., 3 5 May. Lane, G. (2005b), Review of the high frequency Ionospheric Communications Enhanced Profile Analysis and Circuit (ICEPAC) prediction program, paper presented at Ionospheric Effects Symposium, Alexandria, Va., 3 5 May. Lunin, V. Y., and M. M. Woolfson (1993), Mean phase error and the map correlation coefficient, Acta Crystallogr., Sect. D Biol. Crystallogr., 49, McNamara, L. F. (1991), The Ionosphere: Communications, Surveillance, and Direction Finding, Krieger, Malabar, Fla. McNamara, L. F., R. J. Barton, and T. W. Bullet (2006), Analysis of HF signal power observations on two North American circuits, Radio Sci., 41, RS5S38, doi: /2005rs Mudzingwa, C. (2009), A real time HF beacon monitoring station for South Africa, M.S. thesis, Rhodes Univ., Grahamstown, South Africa. Pietrella, M., et al. (2009), Oblique incidence ionospheric soundings over central Europe and their application for testing now casting and long term prediction models, Adv. Space Res., 43, Stocker,A.J.,D.R.Siddle,andE.Warrington(2003),Comparison between the measurement and prediction of HF radio signals propagating along the mid latitude trough, in Ninth International Conference on HF Radio Systems and Techniques, 2003, 493, Walden, M. (2010), A comparison of measurements and propagation simulations for mid latitude HF NVIS links at 5 MHz during sunspot minima, Plextek, Essex, U. K. Zolesi, B., et al. (2008), A new campaign for oblique incidence ionospheric sounding over Europe and its data application, J. Atmos. Sol. Terr. Phys., 70, J. B. Habarulema, L. A. McKinnell, and M. Tshisaphungo, Space Science, South African National Space Agency, PO Box 32, Hermanus 7200, South Africa. (hababosco@gmail.com; lmckinnell@hmo.ac.za; mtshisaphungo@hmo.ac.za) L. Magnus, South African Square Kilometer Array, 3rd Floor, The Park, Park Road, Pinelands 7405, South Africa. (lindsay@ska.ac.za) 7of7