- Email: [email protected]
The variability and IRI2007-predictability of hmF2 over South Africa
Available online at www.sciencedirect.com
Advances in Space Research 52 (2013) 1798–1808 www.elsevier.com/locate/asr
The variability and IRI2007-predictability of hmF2 over South Africa M.C. Mbambo a,b,⇑, Lee-Anne McKinnell a,c, J.B. Habarulema a b
a South African National Space Agency (SANSA), Space Science, P.O. Box 32, Hermanus 7200, South Africa Department of Physics, University of Fort Hare, Private Bag X1314, King William’s Town Rd, Alice 5700, South Africa c Department of Physics and Electronics, Rhodes University, Grahamstown 6140, South Africa
Available online 4 July 2013
Abstract This paper presents an investigation into the variability and predictability of the maximum height of the ionospheric F2 layer, hmF2 over the South African region. Data from three South African stations, namely Madimbo (22.4°S, 26.5°E, dip angle: 61.47°), Grahamstown (33.3°S, 26.5°E, dip angle: 64.08°) and Louisvale (28.5°S, 21.2°E, dip angle: 65.44°) were used in this study. The results indicate that hmF2 shows a larger variability around midnight than during the daytime for all seasons. Monthly median hmF2 values were used in all cases and were compared with predictions from the IRI-2007 model, using the URSI (Union Radio-Scientifique Internationale) coefficient option. The analysis covers the diurnal and seasonal hourly hmF2 values for the selected months and time sectors e.g. January, July, April and October for 2003 and 2005. The time ranges between (03h00–23h00 UT; LT = UT + 2h) representing the local sunrise, midday, sunset and midnight hours. The time covers sunrise, midday, sunrise, and midnight hours (03–06h00 UT, 07–11h00 UT, sunrise 16–18h00 UT and 22–23h00 UT; LT = UT + 2h). The dependence of the results on solar activity levels was also investigated. The IRI2007 predictions follow fairly well the diurnal and seasonal variation patterns of the observed hmF2 values at all the stations. However, the IRI-2007 model overestimates and underestimates the hmF2 value during different months for all the solar activity periods. Ó 2013 COSPAR. Published by Elsevier Ltd. All rights reserved. Keywords: GPS; IRI model; hmF2 parameter; Low-mid latitude ionosphere
1. Introduction This paper presents results from a study undertaken to investigate the variability and predictability of the ionospheric hmF2 parameter over South Africa. The hmF2 parameter is the height at which maximum electron density in the ionosphere occurs (Sethi et al., 2008). The hmF2 parameter is very useful for ionospheric radio-wave propagation studies and for point-to-point communication. A knowledge of hmF2 variability is required in the prediction of High Frequency (HF) radio propagation, as hmF2 can indicate the height at which HF radio frequencies will be ⇑ Corresponding author at: South African National Space Agency (SANSA), Space Science, P.O. Box 32, Hermanus 7200, South Africa. Tel.: +27 28 312 1196; fax: +27 28 312 2039. E-mail addresses: [email protected] (M.C. Mbambo), [email protected] (L.-A. McKinnell), [email protected] (J.B. Habarulema).
reflected and can be used to predict the range of their transmissions (Davies, 1990). Presently, one of the most widely used global empirical models is the International Reference Ionosphere (IRI), which is an ionospheric model based on available data from various sources such as ionosondes, incoherent scatter radars and satellites (Rawer et al., 1981; Bilitza, 1986; Rawer and Bilitza, 1989, 1990). IRI is maintained and revised by an international task group which was established by the Committee on Space Research (COSPAR) and the International Union of Radio Science (URSI). This working group meets annually to update, discuss and plan future improvements to the IRI model. The main aim of the IRI is to provide reliable ionospheric densities, composition and temperature among other ionospheric parameters (Bilitza et al., 1979; Bilitza, 2001). Many other groups have compared their experimental hmF2 values with the IRI model generated values (e.g. Sethi et al., 2004, 2008; Ratovsky et al., 2009; Bertoni
0273-1177/$36.00 Ó 2013 COSPAR. Published by Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.asr.2013.06.022
M.C. Mbambo et al. / Advances in Space Research 52 (2013) 1798–1808
et al., 2006; Zhang et al., 2007; Adewale et al., 2009; Adeniyi and Radicella, 1998; Pandey and Sethi, 1996, and others). Sethi et al. (2008) compared the IRI-2001 model with the observed hmF2 values derived from digital ionosonde measurements at the low-middle latitude station at New Delhi (28.6°N, 77.2°E, dip 42.4°N). They revealed that major discrepancies occur when the IRI overestimates observed hmF2 values during all the seasons for local times from about 06h00 LT till midnight hours, except during the summer period. During other periods observed hmF2 values are compared favorably to the IRI predictions. Similarly, Bertoni et al. (2006) compared IRI-2001 model predictions with ionospheric data measured by digital ionosondes at two Brazilian low-latitude stations, namely Palmas (10.2°S, 48.2°W) and Sa˜o Jose´ dos Campos (23.2°S, 45.9°W). The comparison showed reasonable agreement for both hmF2 and foF2. The authors remarked that some improvements were still necessary to be implemented in order to obtain better predictions. Adewale et al. (2009) compared the monthly means of the ionospheric F2 peak parameters (foF2 and hmF2) over three South African Stations (Grahamstown, Madimbo and Louisvale) with the IRI-2001 model predictions, using both the CCIR and URSI options. Their work showed that the IRI-2001 overestimates observed hmF2 for both quiet and disturbed days. They also showed that foF2 is predicted more accurately by IRI-2001 than hmF2, and on average, the CCIR option performed better than the URSI option when predicting both foF2 and hmF2. While Adewale et al. (2009) compared the South African ionosonde hmF2 values and IRI-2001 predictions for quiet and disturbed periods in 2003, the work presented here also takes into account different solar activity periods. In addition, our analysis was done based on monthly median values for quiet and magnetically disturbed conditions for the entire sunspot cycle. Other groups like Bittencourt and Chryssafidis (1994), Batista et al. (1996) and Shastri et al. (1996) compared their ionospheric data with the IRI-90 (Bilitza, 1990) during different solar activity periods. Batista et al. (1996) used digisonde data that was recorded from 1990 to 1993 at the Cachoeira Paulista station (22.5°S, 45.0°W), and Shastri et al. (1996) compared observed foF2 data from ionosonde measurements for three low-latitude Indian stations namely Delhi (28.6°N, 77.2°E), Ahmedabad (23.0°N, 72.6°E) and Kodaikanal (10.2°N, 77.5°E), with the IRI-90 predictions. Their work showed that the IRI90 model predicted better at different solar activities, except for post-sunset conditions during high solar activity when IRI-90 highly underestimated the observed hmF2. Bittencourt and Chryssafidis (1994) also compared the IRI-90 model predictions with ionospheric values at the Brazilian magnetic equatorial station located at Fortaleza (4.0°S, 38.4°W). Abdu et al. (2006) compared monthly mean variations with IRI representation at Brazilian stations, Sa˜o Luı´s (2.33°S, 44.2°W, dip angle: 0.5°, declination angle: 21°W) and the low latitude station, Cachoeira
1799
Paulista (22.5°S, 315°E, dip angle: 32°). Their work concluded that the daytime hmF2 values were reasonably well represented for Cachoeira Paulista low latitude station in the IRI for both low and high solar flux years, while the nocturnal values were generally underestimated. They also noted that the post-midnight hmF2 enhancement over low latitude station during both 2000 and 1996 are not well represented in the IRI. Firstly, Sobral et al. (2003) compared the monthly averages of foF2 and peak height hmF2 obtained during the solar maximum years, 1978 and 2000 at equatorial and low latitude regions in Brazil with IRI predictions. The equatorial stations are Fortaleza (3°530 S, 38°250 W) and Sa˜o Luı´s (2°200 S, 44°120 W), while Cachoeira Paulista (22°410 S, 45°000 W) was the low latitude station used. Secondly, they compared the electron density profiles obtained by three rocket flights in the Brazilian equatorial region with the IRI predictions at the equatorial stations, Natal (35°140 W, 5°550 S) and Alca´ntara (2°190 S, 44°220 W). Their work showed that the IRI underestimates the equatorial hmF2 during the pre-midnight period. They also showed that the IRI representation for hmF2 over Cachoeira Paulista were in good agreement with the experimental density profile. In addition, they also found good agreement for all local times, between IRI and observed hmF2 values at Cachoeira Paulista for the solar maximum year 2000 and for four months analyzed, e.g February, May, August and November 2000. The objective of this paper is to investigate the predictability of hmF2 under diurnal, seasonal and solar activity variations over the low-mid latitude stations in South Africa. We believe that these results will assist in improving the IRI model over the Southern Hemisphere region. 2. Data used and analysis method The data used in this study was recorded by ionosondes located at three South African stations namely, Madimbo (22.4°S, 26.5°E), Grahamstown (33.3°S, 26.5°E) and Louis-
Fig. 1. Location of the 3 South African ionosonde stations used in this paper.
1800
M.C. Mbambo et al. / Advances in Space Research 52 (2013) 1798–1808
vale (28.5°S, 21.2°E) as shown in Fig. 1. In this work, monthly median values of the hmF2 parameter with an hourly time interval resolution are used. The monthly medians are calculated from the daily hourly values scaled from the ionograms. Firstly we analysed the diurnal hourly values of hmF2 by selecting the months which represent the season of the year, i.e January (summer), July (winter), April (autumn) and October (spring) for 2003. Here, we selected daily hourly data for each month. Then we calculated the monthly median for each hour of the day meaning that there are 24 median values for each month. Secondly, to examine the solar activity variation, the data has been grouped into three time periods of activity, i.e HSA (2000–2002), MSA (2003–2004) and LSA (2005–2006). HSA, MSA, and LSA represent high, medium, and low solar activity levels respectively. Lastly, we analyse the seasonal variation of hmF2 values at local sunrise, midday, sunset and midnight (04h00 UT, midday 10h00 UT, sunset 16h00 UT and midnight 22h00 UT; LT = UT + 2h) respectively. The observed hmF2 values have been com-
pared to the IRI-2007 model (http://nssdc.gsfc.nasa.gov/ space/model/ions/iri.html) predicted values and a statistical analysis has been performed to determine the accuracy of the predictions. The standard deviation is a statistical value used to determine how spread out the data in a sample are, and how close individual data points are to the mean or average value of the sample. In this study, the standard deviation is used for comparing two sets of data (IRI model derived hmF2 and ionosonde hmF2). The standard deviation is expressed in mathematical terms as: vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi u N u1 X ð1Þ S¼t ðxi xÞ2 N i¼1 where x is the sample mean, N is the number of data points, xi represents each data value from i=1 to i= N. The root mean square error (RMSE) has been used here to evaluate the performance of the IRI model when compared to monthly median ionosonde hmF2. The RMSE is expressed in mathematical terms as:
Fig. 2. Comparison of hmF2 variations between ionosonde measurements and IRI-2007 predicted values for: (a) January, (b) April, (c) July and (d) October 2003 over Grahamstown.
M.C. Mbambo et al. / Advances in Space Research 52 (2013) 1798–1808
1801
Fig. 3. Similar to Fig. 2, for Madimbo.
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi u N u 1 X 2 RMSE ¼ t hmF 2obs hmF 2pred N i¼1
ð2Þ
where N is the number of data points, and hmF 2obs and hmF 2pred are the observed and predicted hmF2 values respectively. 3. Results and discussion 3.1. Diurnal variation Monthly median values for ionosonde and IRI-2007 generated hmF2 values were compared for January, April, July and October 2003 over Grahamstown (GRA), Madimbo (MAD) and Louisvale (LOU) ionosonde stations. These particular months were chosen to represent summer, autumn, winter and spring conditions. The seasons cited in this work referred to the Southern hemisphere. Figs. 2– 4(a)–(d) show the hourly averaged hmF2 values for GRA, MAD and LOU stations respectively. The standard
deviation from the observed hmF2 values is indicated by error bars in the plots. Here the observed hmF2 values are shown by solid line, while IRI predicted values are shown by a dashed lines. For Grahamstown (Fig. 2), maximum values of hmF2 for January, April, July and October were observed during the early morning and evening hours for both ionosonde measurements and IRI generated values. On the other hand, both the model and observed ionosonde measurements in Fig. 2 show the lowest values during the period 04h00–06h00 UT. In January and October 2003, minimum monthly median hmF2 values were observed at exactly 04h00 UT as opposed to July 2003 when this minimum occurs at 06h00 UT due to the seasonal variation (McNamara, 1991). For all the months analysed, the IRI-2007 model overpredicts ionosonde hmF2 values. According to Hanson and Patterson (1964) and Kohl and King (1967) the maximum values of hmF2 at night are caused by an increase of upward drifts produced by meridional neutral winds, while during the morning hours the maximum values of hmF2 are caused by an increase in temperature.
1802
M.C. Mbambo et al. / Advances in Space Research 52 (2013) 1798–1808
Fig. 4. Similar to Fig. 2, for Louisvale.
Fig. 3(a)–(d) illustrates that in January and October 2003, both the model and observed ionosonde measurements show the minimum monthly median hmF2 values at 04h00 UT as opposed to April and July 2003 when the minimum values occur at 06h00 UT and 07h00 UT respectively. Fig. 4(a)–(d) shows that the minimum values in January and October 2003 for both model and observed hmF2 values occurred at 04h00 UT as opposed to April and July 2003 when the minimum values occurred at 06h00 UT. For Grahamstown, Madimbo and Louisvale stations the IRI2007 model overpredicts ionosonde hmF2 values for all the months considered. It has been established that the IRI over- or underestimates ionospheric parameters such as foF2 and Total Electron Content (TEC) over South Africa, due to the limited data that has been incorporated into the model over this particular region (McKinnell, 2003; Habarulema et al., 2007, 2009; Oyeyemi et al., 2007; Oyeyemi and Adewale, 2009; Adewale et al., 2009; McKinnell and Oyeyemi, 2009). Bertoni et al. (2006) showed a similar comparison between the observed hmF2 and predicted IRI 2001 values at the Brazilian low latitude stations, namely Sa˜o Jose´ dos
Campos (23.20°S, 45.86°W, dip 38.41°) and Palmas (10.17°S, 48.20°W, dip 10.80°). According to their observations, both predicted and observed hmF2 reached the lowest values between 03h00–06h00 UT during the early morning hours which is more evident in April and July, while in October and January the lowest values occurred between 09h00–12h00 UT during the day. They point out that the model over- or underpredicts hmF2. The observations made by Bertoni et al. (2006) can be compared to the Madimbo station in South Africa, as both Sa˜o Jose´ dos Campos and Madimbo stations are low-mid latitude stations and are located in the Southern Hemisphere at almost similar latitudes. The RMSE method was used to evaluate the hmF2 prediction performance of the IRI-2007 model for selected months in 2003. The RMSE values for the selected months are shown in Fig. 5. The Fig. 5 is constructed by using the average values of Figs. 2–4 respectively. As shown in Fig. 5, January has the highest RMSE values compared to other months, due to poor IRI prediction in summer. Grahamstown station has the lowest RMSE values of all the stations for all the months, except April, while
M.C. Mbambo et al. / Advances in Space Research 52 (2013) 1798–1808
1803
that location than over Madimbo and Louisvale. The lowest RMSE values are seen in July for all three stations, which allows for the conclusion that the IRI provides an improved prediction for winter months than for the summer months. 3.2. Seasonal variation
Fig. 5. A bar graph illustration of RMSE values between measured and IRI-2007 predicted values of hmF2.
Madimbo station has the highest RMSE values for all the months. The low RMSE values over Grahamstown in July simply indicates that the IRI predicts more accurately at
Figs. 6–8(a)–(d) show the comparison of the predicted and observed hmF2 values at local sunrise (04h00 UT), midday (10h00 UT), sunset (16h00 UT) and midnight (22h00 UT). The solid black line indicates the predicted hmF2 values from IRI and the dots indicate the observed hmF2 values. The predicted values were compared with the observed values in order to establish the seasonal performance of the IRI model over these stations for the chosen periods. At local sunrise (04h00 UT), the observed hmF2 shows higher values during winter than summer and reaches a maximum value of 340 km at Louisvale and Grahamstown stations and 310 km at Madimbo. The predicted hmF2 values reach a maximum of approximately 280 km for all the
Fig. 6. Comparison of seasonal variation of observed hmF2 and predicted IRI values over Grahamstown for 2005 at (a) 04h00 UT, (b) 10h00 UT, (c) 16h00 UT and (d) 22h00 UT.
1804
M.C. Mbambo et al. / Advances in Space Research 52 (2013) 1798–1808
Fig. 7. Similar to Fig. 6, for Madimbo.
stations. The predicted IRI hmF2 values do not seem to follow the same trend as the observed hmF2 values during this period. At midday (10h00 UT), both the predicted and observed hmF2 show higher values during equinoxes and summer than during winter. This is because of the seasonal changes due to the relative concentations of atoms and molecular species (McNamara, 1991). For this time, both predicted IRI and observed hmF2 values follow the same trend. The maximum hmF2 value of both the IRI model and the observed hmF2 is around 310 km in January for all the stations. At sunset (16h00 UT), both predicted and observed hmF2 values start at around 280 km for all the stations. The hmF2 values are difficult to predict for this time because the ionosphere is unstable. At midnight (22h00 UT), both predicted and observed hmF2 values are high for all the seasons. However, at both midday (10h00 UT) and midnight (22h00 UT), the observed and predicted hmF2 values are lower during the day than at night over the whole year. This may be because of the influence of meridional winds, since they create a strong upward drift at night and a downward drift during the day (Kohl and King, 1967; Rishbeth and Garriott, 1969).
3.3. Solar activity variation Figs. 9–11(a)–(d) show the variations of mean hmF2 values during low, medium and high solar activity at sunrise (03h00–06h00 UT), noon (07h00–11h00 UT), sunset (16h00–18h00 UT) and midnight (22h00–23h00 UT) hours at Grahamstown, Madimbo and Louisvale stations respectively. The dash and solid lines represent the predicted and observed hmF2 values from different solar activity periods. The analysis covers the mean hmF2 values for HSA (2000– 2002), MSA (2003–2004) and LSA (2005–2006). Firstly, we calculate the monthly medians of the years representing HSA, MSA and LSA. Secondly, the mean of the monthly medians are computed for HSA, MSA and LSA respectively, for selected time intervals representing sunrise (03h00–06h00 UT), noon (07h00–11h00 UT), sunset (16h00–18h00 UT) and midnight (22h00–23h00 UT). The results show that the mean values of hmF2 are generally higher during HSA for all the months at all the locations. According to Stubbe (1964) and Rishbeth (1993) this is caused by temperature variations, which cause hmF2 to increase with increasing solar activity. The height of the F2 peak depends on the temperature profile in the thermo-
M.C. Mbambo et al. / Advances in Space Research 52 (2013) 1798–1808
1805
Fig. 8. Similar to Fig. 6, for Louisvale.
sphere. Stubbe (1964) and Rishbeth (1993) further mention that the increasing solar activity rises the thermospheric temperature, thus causing thermal expansion which lifts the pressure level at which hmF2 is situated. Over Louisvale, the hmF2 values are higher during the night (22h00–23h00 UT) than during the day (07h00– 11h00 UT). According to McNamara (1991) this is to be expected since the morphology of the ionosphere shows it to be high, thin and stable at night. Also hmF2 has lower values during daytime than nighttime for all the months considered. This may be again because of the meridional winds which create a strong upward drift at night and downward drift during day (Kohl and King, 1967; Rishbeth and Garriott, 1969). According to some authors (e.g. Kohl and King, 1967; Rishbeth and Garriott, 1969; Fejer, 1997) the diurnal variation of hmF2 is explained well in terms of ion production rate, temperature variations, thermospheric neutral winds and ExB drifts. Also the effects of winds on hmF2 have been studied by several authors (e.g. Rishbeth, 1967; Miller and Torr, 1987; Miller et al., 1997). Pandey et al. (2003), however presented different results at a low-latitude station for sample noontime electron density profiles during winter and summer seasons
for the Arecibo station (18.4°N, 66.7°W, dip 50°). These authors pointed out that electron density and hmF2 are higher during solar maximum than during the solar minimum period for winter and summer seasons. During sunrise (03h00–06h00 UT) and sunset (16h00– 18h00 UT) hours, hmF2 values are difficult to predict because at this time the ionosphere is not stable (McNamara, 1991). Generally, the predicted IRI hmF2 values overestimate the observed hmF2 values for all the months. During the midnight hours (22h00–23h00 UT), the predicted IRI hmF2 values for MSA underpredicts the ionosonde hmF2 values for MSA from March to April and August to September. According to McKinnell (2002) and Adewale et al. (2009), the over- or underprediction is attributed to the lack of data ingested in the development of the IRI model for the Southern African region. Fig. 11 shows a similar comparison to that of Fig. 9. At midnight hours (22h00–23h00 UT) from May to August, the ionosonde hmF2 values for HSA are higher than the IRI hmF2 values for HSA, while from April to May the ionosonde hmF2 values for LSA are higher than the IRI hmF2 values for LSA.
1806
M.C. Mbambo et al. / Advances in Space Research 52 (2013) 1798–1808
Fig. 9. Mean values of hmF2, determined from observed and IRI predicted values over Grahamstown for HSA, MSA and LSA during (a) local sunrise, (b) noon, (c) sunset and (d) midnight hours.
Fig. 10. Similar to Fig. 9, for Madimbo.
4. Conclusion This paper investigated the diurnal, seasonal and solar activity variation of hmF2 over South Africa during sum-
mer, winter, autumn and spring for 2003 and 2005. The variability of hmF2 during HSA (2000–2002), MSA (2003–2004) and LSA (2005–2006) was also considered. In both cases the ionosonde observations were compared
M.C. Mbambo et al. / Advances in Space Research 52 (2013) 1798–1808
1807
Fig. 11. Similar to Fig. 9 for Louisvale.
Table 1 Monthly RMSE values showing the statistical summary of the prediction performance during different seasons. Season (in 2003)
Grahamstown RMSE (km)
Louisvale RMSE (km)
Madimbo RMSE (km)
Summer (January) Spring (October) Autumn (April) Winter (June)
28.45 26.48 24.37 15.07
33.5 31.22 23.19 21.12
35.61 31.47 26.04 21.79
to the hmF2 values predicted by the IRI-2007 model (URSI option). The results show that the IRI-2007 model follows the diurnal patterns of the observed values of hmF2, especially in winter (July) and autumn (April) for 2003. However, the IRI-2007 overestimates the values of hmF2 for all the seasons in 2003. The statistical summary of the diurnal prediction performance during different seasons is shown in Table 1. This table shows that RMSE values are lower at Grahamstown than at the other two stations (Madimbo and Louisvale) in winter, an indication that the IRI provides an improved prediction during this season for Grahamstown when compared with the other two stations. At winter sunrise, maximum values were observed for all the stations, due to the lack of radiation reaching the lower parts of the ionosphere in the morning (McNamara, 1991). In this case, the observed hmF2 values
were higher than the predicted hmF2 values. Midday hmF2 values were higher during equinoxes and summer than during winter. According to McNamara (1991) and Rishbeth et al. (2000), this is due to a greater intensity of solar radiation reaching the lower parts of the ionosphere during the day. The higher hmF2 values at midnight can be attributed to the ionosphere being high and thin at night (McNamara, 1991). The IRI-2007 model (URSI) underestimates and overestimates the values of hmF2 at different months for all the seasons, during the different solar activity periods considered. Generally, the IRI overpredicts and underpredicts the hmF2 values for different months during different time sectors due to the historic lack of available data in the Southern African region. We noticed that the observed hmF2 values are close to the predicted ones produced by URSI coefficients of the IRI-2007 model at sunrise, noon, sunset and midnight hours, during all solar activity periods. Acknowledgments M.C. Mbambo would like to acknowledge Dr P. Cilliers, Dr V. Xuza, Dr N. Mbatha, V. Tyalimpi and S. Khanyile for the useful comments and encouragement during the process of preparing this paper. We also thank the SPDF/Modelweb team for making the IRI-2007 model available on the website http:// omniweb.gsfc.nasa.gov/vitmo/iri_vitmo.html.
1808
M.C. Mbambo et al. / Advances in Space Research 52 (2013) 1798–1808
References Abdu, M.A., Batista, I., Reinisch, B.W., Sobral, J., Carrasco, A. Equatorial F region evening vertical drift, and peak height, during southern winter months: a comparison of observational data with the IRI descriptions. Adv. Space Res. 37 (5), 1007–1017, 2006. Adeniyi, J.O., Radicella, S.M. Diurnal variation of ionospheric profile parameters B0 and B1 for an equatorial station at low solar activity. J. Atmos. Sol. Terr. Phys. 60 (3), 381–385, 1998. Adewale, A.O., Oyeyemi, E.O., McKinnell, L.A. Comparisons of observed ionospheric F2 peak parameters with IRI-2001 predictions over South Africa. J. Atmos. Sol. Terr. Phys. 71 (2), 273–284, 2009. Adewale, A.O., Oyeyemi, E.O., Ofuase, U.D. Comparison between observed ionospheric foF2 and IRI-2001 predictions over periods of severe geomagnetic activities at Grahamstown. S. Afr. Adv. Space Res. 45 (3), 368–373, 2009. Batista, I.S., Abdu, M.A., De Medeiros, R.T., De Paula, E.R. Comparison between IRI predictions and digisonde measurements at low latitude station. Adv. Space Res. 18 (6), 49–52, 1996. Bertoni, F., Sahai, Y., Lima, W.L.C., Fagundes, P.R., Pillat, V.G., Becker-Guedes, F., Abalde, J.R. IRI-2001 model predictions compared with ionospheric data observed at Brazilian low latitude stations. Ann. Geophys. 24 (8), 2191–2200, 2006. Bilitza, D. International reference ionosphere: recent developments. Radio Sci. 21 (3), 343–346, 1986. Bilitza, D. International Reference Ionosphere 1990, NSSDC 90-22, National Space Science, Data Center, Greenbelt, Md, 1990. Bilitza, D. International reference ionosphere 2000. Radio Sci. 36 (2), 261– 275, 2001. Bilitza, D., Eyfrig, R., Sheikh, N.M. A global model for the height of the F2-peak using M3000 values from the CCIR numerical map. ITU Telecommun. J. 46, 549–553, 1979. Bittencourt, J.A., Chryssafidis, M. On the IRI model predictions for the low-latitude ionosphere. J. Atmos. Terr. Phys. 56 (8), 995–1009, 1994. Davies, K. Ionospheric Radio. Peter Peregrinus Ltd, London, UK, 1990. Fejer, B. The electrodynamics of the low-latitude ionosphere: recent results and future challenges. J. Atmos. Sol. Terr. Phys. 59 (13), 1465– 1482, 1997. Habarulema, J., McKinnell, L.A., Cilliers, P.J. Prediction of global positioning system total electron content using neural networks over South Africa. J. Atmos. Sol. Terr. Phys. 69 (15), 1842–1850, 2007. Habarulema, J.B., McKinnell, L.A., Cilliers, P.J., Opperman, B.D.L. Application of neural networks to South African GPS TEC modelling. Adv. Space Res. 43 (11), 1711–1720, 2009. Hanson, W., Patterson, T. The maintenance of the night-time F-layer. Planet. Space Sci. 12 (10), 979–997, 1964. Kohl, H., King, J. Atmospheric winds between 100 and 700 km and their effects on the ionosphere. J. Atmos. Terr. Phys. 29 (9), 1045–1062, 1967. McKinnell, L. A neural network based ionospheric model for the bottomside electron density profile over Grahamstown, South Africa. Ph.D. Thesis, Rhodes University, Grahamstown, South Africa, 2002. McKinnell, L.A. A neural network based ionospheric model for the bottomside electron density profile over Grahamstown, South Africa. M.Sc. Thesis, Rhodes University, Grahamstown, South Africa, 2003. McKinnell, L.A., Oyeyemi, E.O. Progress towards a new global foF2 model for the international reference ionosphere (IRI). Adv. Space Res. 43 (11), 1770–1775, 2009. McNamara, L.F. The ionosphere: Communications, Surveillance, and Direction Finding. Krieger Publishing Company, Krieger Drive, Malabar, FL 32950, 1991.
Miller, K., Lemon, M., Richards, P. A meridional wind climatology from a fast model for the derivation of meridional winds from the height of the ionospheric F2 region. J. Atmos. Sol. Terr. Phys. 59 (14), 1805– 1822, 1997. Miller, K., Torr, D. A global study of meridional winds in the thermosphere. Adv. Space Res. 7 (10), 299–302, 1987. Oyeyemi, E., McKinnell, L.A., Poole, A.W.V. Neural network-based prediction techniques for global modeling of M (3000) F2 ionospheric parameter. Adv. Space Res. 39 (5), 643–650, 2007. Oyeyemi, E.O., Adewale, A.O. Predictions using a neural network compared with the IRI- 2001 predictions of ionospheric critical frequency observed at Madimbo, South Africa. J. Sci. Res. Develop. 11, 110–121, 2009. Pandey, V.K., Sethi, N.K. Comparison of hmF2 and midday bottomside electron density profile obtained from IRI and incoherent scatter measurements. Adv. Space Res. 18 (6), 33–37, 1996. Pandey, V.K., Sethi, N.K., Mahajan, K.K. Dependence of F2-peak height on solar activity: a study with incoherent scatter measurements. Adv. Space Res. 31 (3), 543–548, 2003. Ratovsky, K.G., Oinats, A.V., Medvedev, A.V. Diurnal and seasonal variations of F2 layer characteristics over Irkutsk during the decrease in solar activity in 2003–2006: Observations and IRI-2001 model predictions. Adv. Space Res. 43 (11), 1806–1811, 2009. Rawer, K., Bilitza, D. Electron density profile description in the international reference ionosphere. J. Atmos. Terr. Phys. 51 (9–10), 781–790, 1989. Rawer, K., Bilitza, D. International reference ionosphere – plasma densities: status 1988. Adv. Space Res. 10 (8), 5–14, 1990. Rawer, K., Lincoln, J.V., Conkright, R.O. International reference ionosphere-IRI 79. World Data Center A for Solar-Terrestrial Physics, Report UAG-82, Boulder, CO, 245p, 1981. Rishbeth, H. The effect of winds on the ionospheric F2-peak. J. Atmos. Terr. Phys. 29 (3), 225–238, 1967. Rishbeth, H. Day-to-day ionospheric variations in a period of high solar activity. J. Atmos. Terr. Phys. 55 (2), 165–171, 1993. Rishbeth, H., Garriott, O. (Eds.), Introduction to Ionospheric Physics, IEEE Transactions on Image Processing, vol. 14. Academic Press, New York, 1969. ISBN: 978-0-12-588940-7. Rishbeth, H., Muller-Wodarg, I.C.F., Zou, L., Fuller-Rowell, T.J., Millward, G.H., Moffett, R.J., Idenden, D.W., Aylward, A.D. Annual and semiannual variations in the ionospheric F2-layer. II. Physical discussion. Ann. Geophys. 18 (8), 945–956, 2000. Sethi, N.K., Dabas, R.S., Sharma, K. Comparison between IRI predictions and digital ionosonde measurements of hmF2 at New Delhi during low and moderate solar activity. J. Atmos. Sol. Terr. Phys. 70 (5), 756–763, 2008. Sethi, N.K., Dabas, R.S., Vohra, V.K. Diurnal and seasonal variations of hmF2 deduced from digital ionosonde over New Delhi and its comparison with IRI 2001. Ann. Geophys. 22 (2), 453–458, 2004. Shastri, S., Aggarwal, S., Sethi, N.K. Performance of IRI model predictions of F-region for Indian latitudes. Adv. Space Res. 18 (6), 41–44, 1996. Sobral, J., Abdu, M., Muralikrishna, P., LaBelle, J., Castilho, V., Zamlutti, C. Rocket and ground-based electron density soundings versus IRI representation. Adv. Space Res. 31 (3), 569–575, 2003. Stubbe, P. Temperature variation at the F-layer maximum during a sunspot cycle. J. Atmos. Terr. Phys. 26 (11), 1055–1068, 1964. Zhang, M.L., Shi, J.K., Wang, X., Shang, S.P., Wu, S.Z. Ionospheric behavior of the F2 peak parameters foF2 and hmF2 at Hainan and comparisons with IRI model predictions. Adv. Space Res. 39 (5), 661– 667, 2007.
Advances in Space Research 52 (2013) 1798–1808 www.elsevier.com/locate/asr
The variability and IRI2007-predictability of hmF2 over South Africa M.C. Mbambo a,b,⇑, Lee-Anne McKinnell a,c, J.B. Habarulema a b
a South African National Space Agency (SANSA), Space Science, P.O. Box 32, Hermanus 7200, South Africa Department of Physics, University of Fort Hare, Private Bag X1314, King William’s Town Rd, Alice 5700, South Africa c Department of Physics and Electronics, Rhodes University, Grahamstown 6140, South Africa
Available online 4 July 2013
Abstract This paper presents an investigation into the variability and predictability of the maximum height of the ionospheric F2 layer, hmF2 over the South African region. Data from three South African stations, namely Madimbo (22.4°S, 26.5°E, dip angle: 61.47°), Grahamstown (33.3°S, 26.5°E, dip angle: 64.08°) and Louisvale (28.5°S, 21.2°E, dip angle: 65.44°) were used in this study. The results indicate that hmF2 shows a larger variability around midnight than during the daytime for all seasons. Monthly median hmF2 values were used in all cases and were compared with predictions from the IRI-2007 model, using the URSI (Union Radio-Scientifique Internationale) coefficient option. The analysis covers the diurnal and seasonal hourly hmF2 values for the selected months and time sectors e.g. January, July, April and October for 2003 and 2005. The time ranges between (03h00–23h00 UT; LT = UT + 2h) representing the local sunrise, midday, sunset and midnight hours. The time covers sunrise, midday, sunrise, and midnight hours (03–06h00 UT, 07–11h00 UT, sunrise 16–18h00 UT and 22–23h00 UT; LT = UT + 2h). The dependence of the results on solar activity levels was also investigated. The IRI2007 predictions follow fairly well the diurnal and seasonal variation patterns of the observed hmF2 values at all the stations. However, the IRI-2007 model overestimates and underestimates the hmF2 value during different months for all the solar activity periods. Ó 2013 COSPAR. Published by Elsevier Ltd. All rights reserved. Keywords: GPS; IRI model; hmF2 parameter; Low-mid latitude ionosphere
1. Introduction This paper presents results from a study undertaken to investigate the variability and predictability of the ionospheric hmF2 parameter over South Africa. The hmF2 parameter is the height at which maximum electron density in the ionosphere occurs (Sethi et al., 2008). The hmF2 parameter is very useful for ionospheric radio-wave propagation studies and for point-to-point communication. A knowledge of hmF2 variability is required in the prediction of High Frequency (HF) radio propagation, as hmF2 can indicate the height at which HF radio frequencies will be ⇑ Corresponding author at: South African National Space Agency (SANSA), Space Science, P.O. Box 32, Hermanus 7200, South Africa. Tel.: +27 28 312 1196; fax: +27 28 312 2039. E-mail addresses: [email protected] (M.C. Mbambo), [email protected] (L.-A. McKinnell), [email protected] (J.B. Habarulema).
reflected and can be used to predict the range of their transmissions (Davies, 1990). Presently, one of the most widely used global empirical models is the International Reference Ionosphere (IRI), which is an ionospheric model based on available data from various sources such as ionosondes, incoherent scatter radars and satellites (Rawer et al., 1981; Bilitza, 1986; Rawer and Bilitza, 1989, 1990). IRI is maintained and revised by an international task group which was established by the Committee on Space Research (COSPAR) and the International Union of Radio Science (URSI). This working group meets annually to update, discuss and plan future improvements to the IRI model. The main aim of the IRI is to provide reliable ionospheric densities, composition and temperature among other ionospheric parameters (Bilitza et al., 1979; Bilitza, 2001). Many other groups have compared their experimental hmF2 values with the IRI model generated values (e.g. Sethi et al., 2004, 2008; Ratovsky et al., 2009; Bertoni
0273-1177/$36.00 Ó 2013 COSPAR. Published by Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.asr.2013.06.022
M.C. Mbambo et al. / Advances in Space Research 52 (2013) 1798–1808
et al., 2006; Zhang et al., 2007; Adewale et al., 2009; Adeniyi and Radicella, 1998; Pandey and Sethi, 1996, and others). Sethi et al. (2008) compared the IRI-2001 model with the observed hmF2 values derived from digital ionosonde measurements at the low-middle latitude station at New Delhi (28.6°N, 77.2°E, dip 42.4°N). They revealed that major discrepancies occur when the IRI overestimates observed hmF2 values during all the seasons for local times from about 06h00 LT till midnight hours, except during the summer period. During other periods observed hmF2 values are compared favorably to the IRI predictions. Similarly, Bertoni et al. (2006) compared IRI-2001 model predictions with ionospheric data measured by digital ionosondes at two Brazilian low-latitude stations, namely Palmas (10.2°S, 48.2°W) and Sa˜o Jose´ dos Campos (23.2°S, 45.9°W). The comparison showed reasonable agreement for both hmF2 and foF2. The authors remarked that some improvements were still necessary to be implemented in order to obtain better predictions. Adewale et al. (2009) compared the monthly means of the ionospheric F2 peak parameters (foF2 and hmF2) over three South African Stations (Grahamstown, Madimbo and Louisvale) with the IRI-2001 model predictions, using both the CCIR and URSI options. Their work showed that the IRI-2001 overestimates observed hmF2 for both quiet and disturbed days. They also showed that foF2 is predicted more accurately by IRI-2001 than hmF2, and on average, the CCIR option performed better than the URSI option when predicting both foF2 and hmF2. While Adewale et al. (2009) compared the South African ionosonde hmF2 values and IRI-2001 predictions for quiet and disturbed periods in 2003, the work presented here also takes into account different solar activity periods. In addition, our analysis was done based on monthly median values for quiet and magnetically disturbed conditions for the entire sunspot cycle. Other groups like Bittencourt and Chryssafidis (1994), Batista et al. (1996) and Shastri et al. (1996) compared their ionospheric data with the IRI-90 (Bilitza, 1990) during different solar activity periods. Batista et al. (1996) used digisonde data that was recorded from 1990 to 1993 at the Cachoeira Paulista station (22.5°S, 45.0°W), and Shastri et al. (1996) compared observed foF2 data from ionosonde measurements for three low-latitude Indian stations namely Delhi (28.6°N, 77.2°E), Ahmedabad (23.0°N, 72.6°E) and Kodaikanal (10.2°N, 77.5°E), with the IRI-90 predictions. Their work showed that the IRI90 model predicted better at different solar activities, except for post-sunset conditions during high solar activity when IRI-90 highly underestimated the observed hmF2. Bittencourt and Chryssafidis (1994) also compared the IRI-90 model predictions with ionospheric values at the Brazilian magnetic equatorial station located at Fortaleza (4.0°S, 38.4°W). Abdu et al. (2006) compared monthly mean variations with IRI representation at Brazilian stations, Sa˜o Luı´s (2.33°S, 44.2°W, dip angle: 0.5°, declination angle: 21°W) and the low latitude station, Cachoeira
1799
Paulista (22.5°S, 315°E, dip angle: 32°). Their work concluded that the daytime hmF2 values were reasonably well represented for Cachoeira Paulista low latitude station in the IRI for both low and high solar flux years, while the nocturnal values were generally underestimated. They also noted that the post-midnight hmF2 enhancement over low latitude station during both 2000 and 1996 are not well represented in the IRI. Firstly, Sobral et al. (2003) compared the monthly averages of foF2 and peak height hmF2 obtained during the solar maximum years, 1978 and 2000 at equatorial and low latitude regions in Brazil with IRI predictions. The equatorial stations are Fortaleza (3°530 S, 38°250 W) and Sa˜o Luı´s (2°200 S, 44°120 W), while Cachoeira Paulista (22°410 S, 45°000 W) was the low latitude station used. Secondly, they compared the electron density profiles obtained by three rocket flights in the Brazilian equatorial region with the IRI predictions at the equatorial stations, Natal (35°140 W, 5°550 S) and Alca´ntara (2°190 S, 44°220 W). Their work showed that the IRI underestimates the equatorial hmF2 during the pre-midnight period. They also showed that the IRI representation for hmF2 over Cachoeira Paulista were in good agreement with the experimental density profile. In addition, they also found good agreement for all local times, between IRI and observed hmF2 values at Cachoeira Paulista for the solar maximum year 2000 and for four months analyzed, e.g February, May, August and November 2000. The objective of this paper is to investigate the predictability of hmF2 under diurnal, seasonal and solar activity variations over the low-mid latitude stations in South Africa. We believe that these results will assist in improving the IRI model over the Southern Hemisphere region. 2. Data used and analysis method The data used in this study was recorded by ionosondes located at three South African stations namely, Madimbo (22.4°S, 26.5°E), Grahamstown (33.3°S, 26.5°E) and Louis-
Fig. 1. Location of the 3 South African ionosonde stations used in this paper.
1800
M.C. Mbambo et al. / Advances in Space Research 52 (2013) 1798–1808
vale (28.5°S, 21.2°E) as shown in Fig. 1. In this work, monthly median values of the hmF2 parameter with an hourly time interval resolution are used. The monthly medians are calculated from the daily hourly values scaled from the ionograms. Firstly we analysed the diurnal hourly values of hmF2 by selecting the months which represent the season of the year, i.e January (summer), July (winter), April (autumn) and October (spring) for 2003. Here, we selected daily hourly data for each month. Then we calculated the monthly median for each hour of the day meaning that there are 24 median values for each month. Secondly, to examine the solar activity variation, the data has been grouped into three time periods of activity, i.e HSA (2000–2002), MSA (2003–2004) and LSA (2005–2006). HSA, MSA, and LSA represent high, medium, and low solar activity levels respectively. Lastly, we analyse the seasonal variation of hmF2 values at local sunrise, midday, sunset and midnight (04h00 UT, midday 10h00 UT, sunset 16h00 UT and midnight 22h00 UT; LT = UT + 2h) respectively. The observed hmF2 values have been com-
pared to the IRI-2007 model (http://nssdc.gsfc.nasa.gov/ space/model/ions/iri.html) predicted values and a statistical analysis has been performed to determine the accuracy of the predictions. The standard deviation is a statistical value used to determine how spread out the data in a sample are, and how close individual data points are to the mean or average value of the sample. In this study, the standard deviation is used for comparing two sets of data (IRI model derived hmF2 and ionosonde hmF2). The standard deviation is expressed in mathematical terms as: vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi u N u1 X ð1Þ S¼t ðxi xÞ2 N i¼1 where x is the sample mean, N is the number of data points, xi represents each data value from i=1 to i= N. The root mean square error (RMSE) has been used here to evaluate the performance of the IRI model when compared to monthly median ionosonde hmF2. The RMSE is expressed in mathematical terms as:
Fig. 2. Comparison of hmF2 variations between ionosonde measurements and IRI-2007 predicted values for: (a) January, (b) April, (c) July and (d) October 2003 over Grahamstown.
M.C. Mbambo et al. / Advances in Space Research 52 (2013) 1798–1808
1801
Fig. 3. Similar to Fig. 2, for Madimbo.
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi u N u 1 X 2 RMSE ¼ t hmF 2obs hmF 2pred N i¼1
ð2Þ
where N is the number of data points, and hmF 2obs and hmF 2pred are the observed and predicted hmF2 values respectively. 3. Results and discussion 3.1. Diurnal variation Monthly median values for ionosonde and IRI-2007 generated hmF2 values were compared for January, April, July and October 2003 over Grahamstown (GRA), Madimbo (MAD) and Louisvale (LOU) ionosonde stations. These particular months were chosen to represent summer, autumn, winter and spring conditions. The seasons cited in this work referred to the Southern hemisphere. Figs. 2– 4(a)–(d) show the hourly averaged hmF2 values for GRA, MAD and LOU stations respectively. The standard
deviation from the observed hmF2 values is indicated by error bars in the plots. Here the observed hmF2 values are shown by solid line, while IRI predicted values are shown by a dashed lines. For Grahamstown (Fig. 2), maximum values of hmF2 for January, April, July and October were observed during the early morning and evening hours for both ionosonde measurements and IRI generated values. On the other hand, both the model and observed ionosonde measurements in Fig. 2 show the lowest values during the period 04h00–06h00 UT. In January and October 2003, minimum monthly median hmF2 values were observed at exactly 04h00 UT as opposed to July 2003 when this minimum occurs at 06h00 UT due to the seasonal variation (McNamara, 1991). For all the months analysed, the IRI-2007 model overpredicts ionosonde hmF2 values. According to Hanson and Patterson (1964) and Kohl and King (1967) the maximum values of hmF2 at night are caused by an increase of upward drifts produced by meridional neutral winds, while during the morning hours the maximum values of hmF2 are caused by an increase in temperature.
1802
M.C. Mbambo et al. / Advances in Space Research 52 (2013) 1798–1808
Fig. 4. Similar to Fig. 2, for Louisvale.
Fig. 3(a)–(d) illustrates that in January and October 2003, both the model and observed ionosonde measurements show the minimum monthly median hmF2 values at 04h00 UT as opposed to April and July 2003 when the minimum values occur at 06h00 UT and 07h00 UT respectively. Fig. 4(a)–(d) shows that the minimum values in January and October 2003 for both model and observed hmF2 values occurred at 04h00 UT as opposed to April and July 2003 when the minimum values occurred at 06h00 UT. For Grahamstown, Madimbo and Louisvale stations the IRI2007 model overpredicts ionosonde hmF2 values for all the months considered. It has been established that the IRI over- or underestimates ionospheric parameters such as foF2 and Total Electron Content (TEC) over South Africa, due to the limited data that has been incorporated into the model over this particular region (McKinnell, 2003; Habarulema et al., 2007, 2009; Oyeyemi et al., 2007; Oyeyemi and Adewale, 2009; Adewale et al., 2009; McKinnell and Oyeyemi, 2009). Bertoni et al. (2006) showed a similar comparison between the observed hmF2 and predicted IRI 2001 values at the Brazilian low latitude stations, namely Sa˜o Jose´ dos
Campos (23.20°S, 45.86°W, dip 38.41°) and Palmas (10.17°S, 48.20°W, dip 10.80°). According to their observations, both predicted and observed hmF2 reached the lowest values between 03h00–06h00 UT during the early morning hours which is more evident in April and July, while in October and January the lowest values occurred between 09h00–12h00 UT during the day. They point out that the model over- or underpredicts hmF2. The observations made by Bertoni et al. (2006) can be compared to the Madimbo station in South Africa, as both Sa˜o Jose´ dos Campos and Madimbo stations are low-mid latitude stations and are located in the Southern Hemisphere at almost similar latitudes. The RMSE method was used to evaluate the hmF2 prediction performance of the IRI-2007 model for selected months in 2003. The RMSE values for the selected months are shown in Fig. 5. The Fig. 5 is constructed by using the average values of Figs. 2–4 respectively. As shown in Fig. 5, January has the highest RMSE values compared to other months, due to poor IRI prediction in summer. Grahamstown station has the lowest RMSE values of all the stations for all the months, except April, while
M.C. Mbambo et al. / Advances in Space Research 52 (2013) 1798–1808
1803
that location than over Madimbo and Louisvale. The lowest RMSE values are seen in July for all three stations, which allows for the conclusion that the IRI provides an improved prediction for winter months than for the summer months. 3.2. Seasonal variation
Fig. 5. A bar graph illustration of RMSE values between measured and IRI-2007 predicted values of hmF2.
Madimbo station has the highest RMSE values for all the months. The low RMSE values over Grahamstown in July simply indicates that the IRI predicts more accurately at
Figs. 6–8(a)–(d) show the comparison of the predicted and observed hmF2 values at local sunrise (04h00 UT), midday (10h00 UT), sunset (16h00 UT) and midnight (22h00 UT). The solid black line indicates the predicted hmF2 values from IRI and the dots indicate the observed hmF2 values. The predicted values were compared with the observed values in order to establish the seasonal performance of the IRI model over these stations for the chosen periods. At local sunrise (04h00 UT), the observed hmF2 shows higher values during winter than summer and reaches a maximum value of 340 km at Louisvale and Grahamstown stations and 310 km at Madimbo. The predicted hmF2 values reach a maximum of approximately 280 km for all the
Fig. 6. Comparison of seasonal variation of observed hmF2 and predicted IRI values over Grahamstown for 2005 at (a) 04h00 UT, (b) 10h00 UT, (c) 16h00 UT and (d) 22h00 UT.
1804
M.C. Mbambo et al. / Advances in Space Research 52 (2013) 1798–1808
Fig. 7. Similar to Fig. 6, for Madimbo.
stations. The predicted IRI hmF2 values do not seem to follow the same trend as the observed hmF2 values during this period. At midday (10h00 UT), both the predicted and observed hmF2 show higher values during equinoxes and summer than during winter. This is because of the seasonal changes due to the relative concentations of atoms and molecular species (McNamara, 1991). For this time, both predicted IRI and observed hmF2 values follow the same trend. The maximum hmF2 value of both the IRI model and the observed hmF2 is around 310 km in January for all the stations. At sunset (16h00 UT), both predicted and observed hmF2 values start at around 280 km for all the stations. The hmF2 values are difficult to predict for this time because the ionosphere is unstable. At midnight (22h00 UT), both predicted and observed hmF2 values are high for all the seasons. However, at both midday (10h00 UT) and midnight (22h00 UT), the observed and predicted hmF2 values are lower during the day than at night over the whole year. This may be because of the influence of meridional winds, since they create a strong upward drift at night and a downward drift during the day (Kohl and King, 1967; Rishbeth and Garriott, 1969).
3.3. Solar activity variation Figs. 9–11(a)–(d) show the variations of mean hmF2 values during low, medium and high solar activity at sunrise (03h00–06h00 UT), noon (07h00–11h00 UT), sunset (16h00–18h00 UT) and midnight (22h00–23h00 UT) hours at Grahamstown, Madimbo and Louisvale stations respectively. The dash and solid lines represent the predicted and observed hmF2 values from different solar activity periods. The analysis covers the mean hmF2 values for HSA (2000– 2002), MSA (2003–2004) and LSA (2005–2006). Firstly, we calculate the monthly medians of the years representing HSA, MSA and LSA. Secondly, the mean of the monthly medians are computed for HSA, MSA and LSA respectively, for selected time intervals representing sunrise (03h00–06h00 UT), noon (07h00–11h00 UT), sunset (16h00–18h00 UT) and midnight (22h00–23h00 UT). The results show that the mean values of hmF2 are generally higher during HSA for all the months at all the locations. According to Stubbe (1964) and Rishbeth (1993) this is caused by temperature variations, which cause hmF2 to increase with increasing solar activity. The height of the F2 peak depends on the temperature profile in the thermo-
M.C. Mbambo et al. / Advances in Space Research 52 (2013) 1798–1808
1805
Fig. 8. Similar to Fig. 6, for Louisvale.
sphere. Stubbe (1964) and Rishbeth (1993) further mention that the increasing solar activity rises the thermospheric temperature, thus causing thermal expansion which lifts the pressure level at which hmF2 is situated. Over Louisvale, the hmF2 values are higher during the night (22h00–23h00 UT) than during the day (07h00– 11h00 UT). According to McNamara (1991) this is to be expected since the morphology of the ionosphere shows it to be high, thin and stable at night. Also hmF2 has lower values during daytime than nighttime for all the months considered. This may be again because of the meridional winds which create a strong upward drift at night and downward drift during day (Kohl and King, 1967; Rishbeth and Garriott, 1969). According to some authors (e.g. Kohl and King, 1967; Rishbeth and Garriott, 1969; Fejer, 1997) the diurnal variation of hmF2 is explained well in terms of ion production rate, temperature variations, thermospheric neutral winds and ExB drifts. Also the effects of winds on hmF2 have been studied by several authors (e.g. Rishbeth, 1967; Miller and Torr, 1987; Miller et al., 1997). Pandey et al. (2003), however presented different results at a low-latitude station for sample noontime electron density profiles during winter and summer seasons
for the Arecibo station (18.4°N, 66.7°W, dip 50°). These authors pointed out that electron density and hmF2 are higher during solar maximum than during the solar minimum period for winter and summer seasons. During sunrise (03h00–06h00 UT) and sunset (16h00– 18h00 UT) hours, hmF2 values are difficult to predict because at this time the ionosphere is not stable (McNamara, 1991). Generally, the predicted IRI hmF2 values overestimate the observed hmF2 values for all the months. During the midnight hours (22h00–23h00 UT), the predicted IRI hmF2 values for MSA underpredicts the ionosonde hmF2 values for MSA from March to April and August to September. According to McKinnell (2002) and Adewale et al. (2009), the over- or underprediction is attributed to the lack of data ingested in the development of the IRI model for the Southern African region. Fig. 11 shows a similar comparison to that of Fig. 9. At midnight hours (22h00–23h00 UT) from May to August, the ionosonde hmF2 values for HSA are higher than the IRI hmF2 values for HSA, while from April to May the ionosonde hmF2 values for LSA are higher than the IRI hmF2 values for LSA.
1806
M.C. Mbambo et al. / Advances in Space Research 52 (2013) 1798–1808
Fig. 9. Mean values of hmF2, determined from observed and IRI predicted values over Grahamstown for HSA, MSA and LSA during (a) local sunrise, (b) noon, (c) sunset and (d) midnight hours.
Fig. 10. Similar to Fig. 9, for Madimbo.
4. Conclusion This paper investigated the diurnal, seasonal and solar activity variation of hmF2 over South Africa during sum-
mer, winter, autumn and spring for 2003 and 2005. The variability of hmF2 during HSA (2000–2002), MSA (2003–2004) and LSA (2005–2006) was also considered. In both cases the ionosonde observations were compared
M.C. Mbambo et al. / Advances in Space Research 52 (2013) 1798–1808
1807
Fig. 11. Similar to Fig. 9 for Louisvale.
Table 1 Monthly RMSE values showing the statistical summary of the prediction performance during different seasons. Season (in 2003)
Grahamstown RMSE (km)
Louisvale RMSE (km)
Madimbo RMSE (km)
Summer (January) Spring (October) Autumn (April) Winter (June)
28.45 26.48 24.37 15.07
33.5 31.22 23.19 21.12
35.61 31.47 26.04 21.79
to the hmF2 values predicted by the IRI-2007 model (URSI option). The results show that the IRI-2007 model follows the diurnal patterns of the observed values of hmF2, especially in winter (July) and autumn (April) for 2003. However, the IRI-2007 overestimates the values of hmF2 for all the seasons in 2003. The statistical summary of the diurnal prediction performance during different seasons is shown in Table 1. This table shows that RMSE values are lower at Grahamstown than at the other two stations (Madimbo and Louisvale) in winter, an indication that the IRI provides an improved prediction during this season for Grahamstown when compared with the other two stations. At winter sunrise, maximum values were observed for all the stations, due to the lack of radiation reaching the lower parts of the ionosphere in the morning (McNamara, 1991). In this case, the observed hmF2 values
were higher than the predicted hmF2 values. Midday hmF2 values were higher during equinoxes and summer than during winter. According to McNamara (1991) and Rishbeth et al. (2000), this is due to a greater intensity of solar radiation reaching the lower parts of the ionosphere during the day. The higher hmF2 values at midnight can be attributed to the ionosphere being high and thin at night (McNamara, 1991). The IRI-2007 model (URSI) underestimates and overestimates the values of hmF2 at different months for all the seasons, during the different solar activity periods considered. Generally, the IRI overpredicts and underpredicts the hmF2 values for different months during different time sectors due to the historic lack of available data in the Southern African region. We noticed that the observed hmF2 values are close to the predicted ones produced by URSI coefficients of the IRI-2007 model at sunrise, noon, sunset and midnight hours, during all solar activity periods. Acknowledgments M.C. Mbambo would like to acknowledge Dr P. Cilliers, Dr V. Xuza, Dr N. Mbatha, V. Tyalimpi and S. Khanyile for the useful comments and encouragement during the process of preparing this paper. We also thank the SPDF/Modelweb team for making the IRI-2007 model available on the website http:// omniweb.gsfc.nasa.gov/vitmo/iri_vitmo.html.
1808
M.C. Mbambo et al. / Advances in Space Research 52 (2013) 1798–1808
References Abdu, M.A., Batista, I., Reinisch, B.W., Sobral, J., Carrasco, A. Equatorial F region evening vertical drift, and peak height, during southern winter months: a comparison of observational data with the IRI descriptions. Adv. Space Res. 37 (5), 1007–1017, 2006. Adeniyi, J.O., Radicella, S.M. Diurnal variation of ionospheric profile parameters B0 and B1 for an equatorial station at low solar activity. J. Atmos. Sol. Terr. Phys. 60 (3), 381–385, 1998. Adewale, A.O., Oyeyemi, E.O., McKinnell, L.A. Comparisons of observed ionospheric F2 peak parameters with IRI-2001 predictions over South Africa. J. Atmos. Sol. Terr. Phys. 71 (2), 273–284, 2009. Adewale, A.O., Oyeyemi, E.O., Ofuase, U.D. Comparison between observed ionospheric foF2 and IRI-2001 predictions over periods of severe geomagnetic activities at Grahamstown. S. Afr. Adv. Space Res. 45 (3), 368–373, 2009. Batista, I.S., Abdu, M.A., De Medeiros, R.T., De Paula, E.R. Comparison between IRI predictions and digisonde measurements at low latitude station. Adv. Space Res. 18 (6), 49–52, 1996. Bertoni, F., Sahai, Y., Lima, W.L.C., Fagundes, P.R., Pillat, V.G., Becker-Guedes, F., Abalde, J.R. IRI-2001 model predictions compared with ionospheric data observed at Brazilian low latitude stations. Ann. Geophys. 24 (8), 2191–2200, 2006. Bilitza, D. International reference ionosphere: recent developments. Radio Sci. 21 (3), 343–346, 1986. Bilitza, D. International Reference Ionosphere 1990, NSSDC 90-22, National Space Science, Data Center, Greenbelt, Md, 1990. Bilitza, D. International reference ionosphere 2000. Radio Sci. 36 (2), 261– 275, 2001. Bilitza, D., Eyfrig, R., Sheikh, N.M. A global model for the height of the F2-peak using M3000 values from the CCIR numerical map. ITU Telecommun. J. 46, 549–553, 1979. Bittencourt, J.A., Chryssafidis, M. On the IRI model predictions for the low-latitude ionosphere. J. Atmos. Terr. Phys. 56 (8), 995–1009, 1994. Davies, K. Ionospheric Radio. Peter Peregrinus Ltd, London, UK, 1990. Fejer, B. The electrodynamics of the low-latitude ionosphere: recent results and future challenges. J. Atmos. Sol. Terr. Phys. 59 (13), 1465– 1482, 1997. Habarulema, J., McKinnell, L.A., Cilliers, P.J. Prediction of global positioning system total electron content using neural networks over South Africa. J. Atmos. Sol. Terr. Phys. 69 (15), 1842–1850, 2007. Habarulema, J.B., McKinnell, L.A., Cilliers, P.J., Opperman, B.D.L. Application of neural networks to South African GPS TEC modelling. Adv. Space Res. 43 (11), 1711–1720, 2009. Hanson, W., Patterson, T. The maintenance of the night-time F-layer. Planet. Space Sci. 12 (10), 979–997, 1964. Kohl, H., King, J. Atmospheric winds between 100 and 700 km and their effects on the ionosphere. J. Atmos. Terr. Phys. 29 (9), 1045–1062, 1967. McKinnell, L. A neural network based ionospheric model for the bottomside electron density profile over Grahamstown, South Africa. Ph.D. Thesis, Rhodes University, Grahamstown, South Africa, 2002. McKinnell, L.A. A neural network based ionospheric model for the bottomside electron density profile over Grahamstown, South Africa. M.Sc. Thesis, Rhodes University, Grahamstown, South Africa, 2003. McKinnell, L.A., Oyeyemi, E.O. Progress towards a new global foF2 model for the international reference ionosphere (IRI). Adv. Space Res. 43 (11), 1770–1775, 2009. McNamara, L.F. The ionosphere: Communications, Surveillance, and Direction Finding. Krieger Publishing Company, Krieger Drive, Malabar, FL 32950, 1991.
Miller, K., Lemon, M., Richards, P. A meridional wind climatology from a fast model for the derivation of meridional winds from the height of the ionospheric F2 region. J. Atmos. Sol. Terr. Phys. 59 (14), 1805– 1822, 1997. Miller, K., Torr, D. A global study of meridional winds in the thermosphere. Adv. Space Res. 7 (10), 299–302, 1987. Oyeyemi, E., McKinnell, L.A., Poole, A.W.V. Neural network-based prediction techniques for global modeling of M (3000) F2 ionospheric parameter. Adv. Space Res. 39 (5), 643–650, 2007. Oyeyemi, E.O., Adewale, A.O. Predictions using a neural network compared with the IRI- 2001 predictions of ionospheric critical frequency observed at Madimbo, South Africa. J. Sci. Res. Develop. 11, 110–121, 2009. Pandey, V.K., Sethi, N.K. Comparison of hmF2 and midday bottomside electron density profile obtained from IRI and incoherent scatter measurements. Adv. Space Res. 18 (6), 33–37, 1996. Pandey, V.K., Sethi, N.K., Mahajan, K.K. Dependence of F2-peak height on solar activity: a study with incoherent scatter measurements. Adv. Space Res. 31 (3), 543–548, 2003. Ratovsky, K.G., Oinats, A.V., Medvedev, A.V. Diurnal and seasonal variations of F2 layer characteristics over Irkutsk during the decrease in solar activity in 2003–2006: Observations and IRI-2001 model predictions. Adv. Space Res. 43 (11), 1806–1811, 2009. Rawer, K., Bilitza, D. Electron density profile description in the international reference ionosphere. J. Atmos. Terr. Phys. 51 (9–10), 781–790, 1989. Rawer, K., Bilitza, D. International reference ionosphere – plasma densities: status 1988. Adv. Space Res. 10 (8), 5–14, 1990. Rawer, K., Lincoln, J.V., Conkright, R.O. International reference ionosphere-IRI 79. World Data Center A for Solar-Terrestrial Physics, Report UAG-82, Boulder, CO, 245p, 1981. Rishbeth, H. The effect of winds on the ionospheric F2-peak. J. Atmos. Terr. Phys. 29 (3), 225–238, 1967. Rishbeth, H. Day-to-day ionospheric variations in a period of high solar activity. J. Atmos. Terr. Phys. 55 (2), 165–171, 1993. Rishbeth, H., Garriott, O. (Eds.), Introduction to Ionospheric Physics, IEEE Transactions on Image Processing, vol. 14. Academic Press, New York, 1969. ISBN: 978-0-12-588940-7. Rishbeth, H., Muller-Wodarg, I.C.F., Zou, L., Fuller-Rowell, T.J., Millward, G.H., Moffett, R.J., Idenden, D.W., Aylward, A.D. Annual and semiannual variations in the ionospheric F2-layer. II. Physical discussion. Ann. Geophys. 18 (8), 945–956, 2000. Sethi, N.K., Dabas, R.S., Sharma, K. Comparison between IRI predictions and digital ionosonde measurements of hmF2 at New Delhi during low and moderate solar activity. J. Atmos. Sol. Terr. Phys. 70 (5), 756–763, 2008. Sethi, N.K., Dabas, R.S., Vohra, V.K. Diurnal and seasonal variations of hmF2 deduced from digital ionosonde over New Delhi and its comparison with IRI 2001. Ann. Geophys. 22 (2), 453–458, 2004. Shastri, S., Aggarwal, S., Sethi, N.K. Performance of IRI model predictions of F-region for Indian latitudes. Adv. Space Res. 18 (6), 41–44, 1996. Sobral, J., Abdu, M., Muralikrishna, P., LaBelle, J., Castilho, V., Zamlutti, C. Rocket and ground-based electron density soundings versus IRI representation. Adv. Space Res. 31 (3), 569–575, 2003. Stubbe, P. Temperature variation at the F-layer maximum during a sunspot cycle. J. Atmos. Terr. Phys. 26 (11), 1055–1068, 1964. Zhang, M.L., Shi, J.K., Wang, X., Shang, S.P., Wu, S.Z. Ionospheric behavior of the F2 peak parameters foF2 and hmF2 at Hainan and comparisons with IRI model predictions. Adv. Space Res. 39 (5), 661– 667, 2007.