Introduction

In the middle to late 1930s, ionospheric irregularities and the manner in which their electrodynamic mechanisms affected ionospheric behaviors began to attract the interest of many researchers (Abdu et al. 1981a, b, 1998, 2009; Booker and Wells 1938; Bowman 1974, 1990; Chandra and Rastogi 1970; Chou and Kuo 1996; de Jesus et al. 2013; Ossakow 1981; Xiong et al. 2012). Ionospheric irregularities appear as scattered echoes in high-frequency (HF) band ionograms that are known as spread-F events. Spread-Fs can manifest as frequency spread-Fs (FSF) that are broadened traces that mark reflections from the ionosphere along the frequency axis, or as range spread-Fs (RSF) that are along the vertical height axis. Many ground-based instruments (optical, ionosondes, and radar) and space-borne platforms (rockets and satellites) have been employed to explore the spread-F phenomenon over the past seven decades. These efforts have deepened our knowledge on spread-Fs showing that they vary with respect to latitude, local time, season, and solar and magnetic activity (Alfonsi et al. 2013; Banola et al. 2005; Chou and Kuo 1996; Deng et al. 2013; Huang et al. 1993; Scherliess and Fejer 1999). Different mechanisms have been proposed to explain spread-F occurrences and their development (Bowman 1990; Fejer et al. 1999; Fukao et al. 2004); among these, the primary mechanism in equatorial regions is the generalized Rayleigh–Taylor (R–T) instability mechanism. The R–T instability mechanism suggests that pre-reversal electric field enhancements (PRE) during the evening cause a rapid uplift of the ionosphere’s F-layer (Fejer et al. 1999; Fukao et al. 2004; Manju et al. 2007; Sukanta et al. 2017; Xiong et al. 2012; Upadhayaya and Gupta 2014). Relationships between spread-Fs and other ionospheric parameters, particularly the F2-layer (foF2) and h′F variations with the occurrence of spread-Fs, have also been statistically examined (Rungraengwajiake et al. 2013; Joshi et al. 2013; Madhav Haridas et al. 2013; de Abreu et al. 2014a, b, c, 2017; Abadi et al. 2015; Manju and Madhav Haridas 2015; Smith et al. 2015; Liu and Shen 2017). In addition, the effects of seasonal, solar, and magnetic activity variabilities on the h′F threshold have also been investigated (Manju et al. 2007; Manju and Madhav Haridas 2015; Madhav Haridas et al. 2013; Stoneback et al. 2011; Narayanan et al. 2014, 2017).

Devasia et al. (2002) first introduced the concept of threshold height (h′Fc) as a critical parameter controlling the day-to-day equatorial spread-F (ESF) variability. Past studies have revealed the dependence of the h′Fc on seasonal variations and solar and magnetic activity for the occurrence of ESFs and found the occurrences to be irrespective of the magnitude and polarity of meridional winds (Jyoti et al. 2004; Manju et al. 2007). Rungraengwajiake et al. (2013) presented a comparative study of the correlation between h′F and RSF occurrences in Thailand, and the results showed that high RSF occurrences mostly happened during equinoctial months that corresponded to rapid increases in the monthly mean h′F after sunset. Joshi et al. (2013) found that the h′F plays a key role in determining the R–T instability growth rate. Madhav Haridas et al. (2013) presented the effects of seasonal and solar activity variations of the h′Fc on ESF occurrences in India and found that substantial increases in the h′Fc varied with magnetic activity during every season.

Similar studies in Brazil have been presented (de Abreu et al. 2014a, b, c) to show that the occurrence of ESFs are closely related to daily variations of the h′F near the equator. During periods of low solar activity (LSA), the 250 km h′F altitude acted as the h′Fc for the generation of spread-Fs, while the 300 km h′Fc was during periods of high solar activity (HSA). An investigation using measurements from multiple instruments over the American sector showed that spread-Fs were often observed the nights before and during storms near the equator, in which the foF2 was less than 8 MHz and the h′F was lower than 300 km (de Abreu et al. 2017).

Abadi et al. (2015) studied the influences of the h′F on the latitudinal extension of ionospheric irregularities in Southeast Asia. Their results suggested that the latitudinal extension of plasma bubbles was mainly controlled by the PRE magnitude and h′F peak values during the initial phases of the ESF. Manju and Madhav Haridas (2015) investigated the h′Fc for the occurrences of ESFs during equinoxes and showed that the equinoctial asymmetry of the h′Fc increases with solar activity. Aside from the studies mentioned above, there are few reports that consider the effect of the foF2 threshold on the generation of spread-F events. Liu and Shen (2017) conducted a case study during a severe geomagnetic storm near 120°E in China and showed that the spread-F was suppressed near Sanya and Wuhan during the storm’s main phase when the frequency spread over 14 MHz, and the suppression was sustained for several hours. This helped us to understand the possible onset causes of the day-to-day spread-F variability.

Stoneback et al. (2011) investigated the local time distribution of meridional (vertical) drifts during the prolonged solar minimum. They found that the downward drifts across sunset and the upward drifts across midnight were also consistent with the delay in the appearance of ionospheric irregularities after midnight. Narayanan et al. (2014) studied the relationship between the occurrence of satellite traces (STs) in ionograms and the formation of ESFs using observations from an Indian dip equatorial station during solar minimum conditions. They found that the ST occurred later in the night as well implying that the PRE was not the cause of the ST during these times. Additionally, they also found that the STs were not followed by ESFs in about 30% of the cases indicating that large-scale wave-like structures (LSWS) do not trigger ESFs on all occasions. Narayanan et al. (2017) also found that the plasma bubbles were generated without strong PREs when the ion-neutral collision frequencies possibly dropped significantly during the unusually low solar activity conditions of 2008. Abdu et al. (2006) found that the existence of significant planetary wave (PW) influences on plasma parameters at E- and F-region heights over the equatorial latitudes using airglow, radar, and ionospheric sounding observations. A direct consequence of the PW scale oscillations in the evening electric field is its role in the quiet time day-to-day variability of the ESF/plasma bubble occurrences and intensities.

We limited our focus to spread-F occurrences and their relationships with foF2 and h′F that affected spread-F occurrences during a complete solar cycle in the low- and mid-latitudes over China. The International Reference Ionosphere-2012 (IRI-2012) model includes the monthly mean spread-F occurrences for predicting in the Brazilian longitude sector but not for Chinese sector. Therefore, the studies of spread-F occurrence statistics in China are part of an on-going effort to develop the spread-F occurrence prediction abilities to improve the IRI model. In the present study, we focused on the characteristics and correlations between spread-F occurrences and the foF2 and h′F. Furthermore, we also present the thresholds of the foF2 as they relate to the generation of FSFs.

Data and analysis

The China Research Institute of Radio-wave Propagation (CRIRP) constructed and operated a network of long-running ionospheric observation sites that cover mainland China. In this study, we extracted simultaneous spread-F data from four digital ionosondes located at Haikou (HK) (20°N, 110.34°E), Guangzhou (GZ) (23.14°N, 113.36°E), Beijing (BJ) (40.11°N, 116.28°E), and Changchun (CC) (43.84°N, 125.28°E). In addition, we also determined the data characteristic of the foF2 and h′F at these sites to reveal possible correlations between spread-F occurrences and the foF2 and h′F. No data were recorded in December 1997 and from May to December 1999 at CC, because the ionosonde was being repaired. The observational site details are shown in Table 1.

Table 1 Details of the digital ionosonde sites used in the investigation
Full size table

The HK and GZ sites lie near the north crest of the equatorial ionization anomaly (EIA) zone. The EIA zone is where the fountain effect phenomena and the equatorial electrojet often interact resulting in complicated ionospheric physical processes. BJ and CC are located at the mid-latitudes in China. According to previous studies, ionospheric irregularities greatly depend on solar activity, local time, season, latitude and longitude, and geomagnetic disturbances (Abdu et al. 1981a, b, 1983, 1998, 2009; Booker and Wells 1938; Bowman 1974; Chandra and Rastogi 1970; Maruyama 1988; Xiong et al. 2012). To discuss the correlations between spread-Fs and solar and geomagnetic activities, we show the monthly mean 10.7 cm radio flux (F10.7) and ap index during the 23rd solar cycle in Fig. 1 that covers the epochs of the LSA and HSA. We used a 3-hourly ap index to identify geomagnetically quiet and disturbed days. If the maximum value of the 3-hourly ap index for a day was greater than 12, the day was considered as a disturbed day (Narayanan et al. 2017). Figure 2 shows the daily max ap indices from 2000 to 2005. Further, it can be seen from the figure that there were more geomagnetically disturbed days during the vernal equinox and autumn equinoxes in 2001 and 2002.

Fig. 1
figure 1

Monthly averaged 10.7 cm solar flux (F10.7) (y axis: F10.7/sfu) and ap index from 1997 to 2008 denoting the solar activity

Full size image
Fig. 2
figure 2

Daily max ap indices from 2000 to 2005

Full size image

Ionogram data were collected using type TYC-1 ionosondes, which are designed and manufactured by the CRIRP (Xu et al. 2001). Ionograms were recorded at 1-h intervals for a frequency range from 1 to 32 MHz. We distinguished two types of spread-F, FSF, and RSF for detailed study. We used the percentage of spread-F occurrences to describe the spread-F statistical features, which is defined as follows:

$${\text{P}}\left( y, m, h \right) = \frac{{n \left( {y, m, h} \right)}}{{N\left( {y,m,h} \right)}} \times 100{\text{\% }}$$
(1)

where y, m, and h represent the year, month, and local time (LT), respectively; n is the number of spread-F occurrences that appear at the same local time but during different days of a single month, and N is the total number of days for a given year and local time. Spread-Fs typically appeared after sunset and lasted until the subsequent sunrise; thus, the percentage of spread-F occurrences from 18:00 LT to 06:00 LT is the topic of interest in this study. Occurrences of FSF and RSF were compared with monthly medians of the foF2 and h′F to find the correlations between foF2 and h′F for the generation of spread-Fs. The FSF, RSF, foF2, and h′F were differentiated by manually analyzing the ionograms. The foF2 and h′F can sometimes be measured, but sometimes cannot be obtained when a spread-F occurs. The foF2 and h′F cannot be obtained during a strong spread-F (SSF). SSFs are a type of spread-F that can be identified when there is strong diffusion on the frequency and height axis of an ionogram. Figure 3 shows a SSF event in Haikou on March 26, 2012. The observations presented in this manuscript contain data when the foF2 and h′F values were reliably scaled during a spread-F. To examine their seasonal variations, we grouped the data into the following four seasonal bins: summer (May, June, July and August), vernal equinox (March and April), autumn equinox (September and October), and winter (January, February, November and December) (Maruyama and Matuura 1984; Maruyama et al. 2009; Sripathi et al. 2011; Xiao and Zhang 2001).

Fig. 3
figure 3

A SSF event in Haikou on March 26, 2012

Full size image

Results and discussion

Nocturnal, seasonal, and solar activity variations on spread-F occurrences

The monthly mean of the FSF occurrence rates varied with local time and are presented separately in Fig. 4 for Haikou, Guangzhou, Beijing, and Changchun. It can be found that the FSF occurrences frequently appeared after midnight. Also, the FSF occurrences observed at different sites exhibited distinct local time distribution patterns. Previous studies have also observed this trend (Zhang et al. 2015; de Jesus et al. 2010, 2012, 2016). The FSF occurrence rates at HK, BJ and CC were higher than GZ. The maximum FSF occurrence rate was ~ 80% and occurred in July 1997 at HK, in August 2008 at BJ and in June 2006 at CC. The LSA yielded high FSF occurrence percentages at all four sites. The relationship between the FSF and solar activity was approximate to a negative correlation. The seasonal variation of the FSF occurrence rates observed at the four sites is shown in Fig. 5a–d. We found that FSFs occurred mostly during the summer at HK and the occurrence rate was lower between 1999 and 2002. FSF occurrence rates were higher during the autumn equinox than during the vernal equinox between 2000 and 2001 at HK. FSFs occurred mostly during the summer at GZ, however, scarcely occurred in 2002 and 2008. Statistically, the FSFs started at approximately 21:00 LT and lasted until 05:00 LT at HZ and CC. However, FSFs started at about 23:00 LT and lasted until 05:00 LT at GZ and BJ, with post-midnight FSFs as the most commonly observed.

Fig. 4
figure 4

Monthly mean FSF occurrence percentages at the four sites

Full size image
Fig. 5
figure 5

Seasonal variation of the FSF occurrences observed at HK (a), GZ (b), BJ (c), and CC (d)

Full size image

Figure 6 shows variations in the average RSF occurrence rates at the four sites. The RSF occurrence rate was much larger than the FSF occurrence rate at GZ; however, the rates were smaller than the FSF occurrence rates at BJ and CC. RSF occurrence rates increased with an increase in solar activity at HK, but not at the other three sites. The maximum RSF occurrence rate was higher than 80% in June 2006 and July 2007 at GZ. Figure 7 shows the seasonal RSF occurrence rate variations at the four sites. RSFs mostly occurred in the vernal equinox and autumn equinox months during HSA years at HK and GZ. These observations revealed that the RSF occurrence rate from 2000 to 2002 at HK and GZ were possibly affected by the geomagnetic activity according to Fig. 2. During the solar maximum period between 2000 and 2002, RSFs appeared earlier than during other periods, with a maximum RSF occurrence rate occurring between 21:00 LT and 01:00 LT at HK and GZ. Different from the FSF occurrences, higher RSF occurrence rates mostly occurred during the winter months at BJ and CC. Previous studies have emphasized that FSF events are well correlated with bottom-side layers, while RSFs are closely correlated with plumes. Additionally, the RSF occurrence rate reaches its maximum before midnight during HSA at low latitude, whereas that of an FSF reaches a maximum after midnight (Liu et al. 2004a, b; Chen et al. 2006; Aarons et al. 1994; Hu et al. 2004). This regular pattern was also observed at the four sites in China.

Fig. 6
figure 6

Same as Fig. 2, but for the RSF occurrences

Full size image
Fig. 7
figure 7

Seasonal variation of the RSF occurrences at HK (a), GZ (b), BJ (c), and CC (d)

Full size image

Abdu et al. (2003) showed that RSF events are associated with developed or developing plasma bubble events, while FSF events are associated with narrow-spectrum irregularities that occur near the peak of the F-layer. These results suggest that the upward velocity of plasma bubbles have a strong seasonal connection with the maximum values observed during the summer. Variations of FSF and RSF except for those during the 2000–2002 solar maximum period are mainly consistent with these studies. Rungraengwajiake et al. (2013) showed that FSF events appear later than RSF events on average and that FSFs remain until morning, while RSFs almost disappear by around 04:00 LT. The results shown in Figs. 5 and 7 are slightly different, which may be partly attributed to the effects of geomagnetic activity. Figure 2 shows the geomagnetic activity during the equinoxes in 2001 and 2002. It is possible these activities caused the RSFs to occur mainly during equinoxes at HK and GZ in 2001 and 2002. The peak FSF occurrence rate appeared later at GZ than at HK, which is well correlated with the manner in which fresh bubbles start from the latter station and then expand to high latitudes. The average FSF occurrence percentage mostly peaks from 24:00 LT to 02:00 LT at HK and from 03:00 LT to 05:00 LT at GZ. The average RSF occurrence percentages mostly peaked from 21:00 LT to 23:00 LT at HK and from 24:00 LT to 02:00 LT at GZ during periods of HSA. Meanwhile, RSF occurrence rates were higher at HK and GZ than those at BJ and CC; FSF occurrence rates were higher at HK, BJ, and CC than at GZ. These results support the hypothesis that solar and geomagnetic activity affects seasonal and longitudinal variations of spread-Fs.

Liu and Shen (2017) found that the disturbance of electric fields could also contribute to the occurrence of spread-Fs, especially at low-latitude stations. The disturbed electric fields and the disturbance winds are also the probable factors that promote the spread-F along with the gravity-driven R–T instability. In addition, the electric field disturbances can also generate spread-Fs through R–T instability only (de Jesus et al. 2010; Wang et al. 2014; Wan and Xu 2014; Mo et al. 2017). The disturbance of the dynamo driven by enhanced global thermospheric circulation resulting from energy input at high latitudes is another factor for promoting spread-Fs (de Jesus et al. 2010; Liu and Shen 2017). Therefore, it can be seen that there are many possible mechanisms for spread-F occurrences, and more in-depth analysis is needed.

Nocturnal, seasonal, and solar activity variations on foF2 and h′F

In Fig. 8 we showed local time and the variations in solar activity in the monthly median foF2 data from the 23rd solar cycle. At lower latitudes, a higher magnitude foF2 was sustained until midnight. In addition, another morphological feature of the monthly medians is the typical post-sunset peak values. Between 1998 and 2005, foF2 variations showed dual-peak patterns at HK and GZ that reached a minimum during the summer and a maximum during the spring and winter. Additionally, wintertime monthly medians of the foF2 were higher during the spring in 1998, but this result is inverted between 2003 and 2005. Figure 9 shows the seasonal variations of the averaged foF2 monthly median data at all four sites. The medians reached their peak magnitudes between 18:00 and 19:00 LT. In addition, the medians were mostly higher during equinox seasons at HK and GZ; however, they were mostly higher during the summer at BJ and CC. The highest foF2 medians were ~ 18 MHz and occurred from 18:00 LT to 24:00 LT at HK and GZ during periods of maximum solar activity. The minimal medians occurred before dawn from around 03:00–05:00 LT. The post-midnight collapse of the foF2 usually occurred more often at low latitudes than mid-latitudes.

Fig. 8
figure 8

Variation of the monthly median foF2 at the four sites between 1997 and 2008

Full size image
Fig. 9
figure 9

Seasonal variation of the monthly median foF2 at HK (a), GZ (b), BJ (c), and CC (d)

Full size image

Abdu et al. (1983) proposed that the h′F parameter may be a possible factor involved in the occurrence and variation of spread-Fs. Figure 10 shows the h′F monthly median data at all four sites, thus demonstrating that monthly medians were higher at HK and GZ than at BJ and CC. The peak median h′F values occurred before midnight during the summer in HSA at HK and GZ; however, the peak value onset time was later at high latitudes. During periods of HSA, monthly medians increase. Figure 11 shows the seasonal variation of the average h′F monthly median at the four sites, which is quite different from the foF2. The maximum h′F values occurred from 21:00 LT to 01:00 LT during summer months at HK and GZ; otherwise, they appeared at or before midnight from 2000 to 2002.

Fig. 10
figure 10

Same as Fig. 6, but for monthly medians of the h′F

Full size image
Fig. 11
figure 11

Seasonal variation of the monthly medians of the h′F at HK (a), GZ (b), BJ (c), and CC (d)

Full size image

The possible foF2 threshold for FSFs and the relationship between the h′F and RSF

The correlations between spread-F occurrence and the foF2 and h′F magnitudes are discussed in this section. Figures 12 and 13 show the post-sunset foF2 and h′F variations compared with the normalized spread-F occurrence rates at the four sites. In order to analyze the correlation between the spread-F occurrence rate and the foF2 and h′F, the normalized probability was used. The normalized spread-F occurrence rate is defined as follows:

$$p_{i} = \frac{{m}_{i} }{{\mathop \sum \nolimits_{i} {m}_{i} }}$$
(2)
$$\mathop \sum \limits_{i} {p}_{i} = 1$$
(3)

where p is the normalized FSF or RSF occurrence rate, m i is the number of FSF or RSF event occurrences when the foF2 or h′F is within a certain interval. We used 0.2 MHz and 5 km as the sampling intervals for the foF2 and h′F. The summation of m i is the total number of FSF or RSF event occurrences. We applied the polynomial fitting method during the relationship analysis between the foF2 and h′F and the spread-F occurrence rates. We found that the foF2 and FSF occurrences satisfy the linear relationship shown in Fig. 12, and the h′F and RSF occurrences are similar to parabolic relationship shown in Fig. 13. The red point is the sample value. The blue lines are a fitting line or curve. The FSF occurrence rates increased with a decrease in foF2 at each site, and the foF2 values ranged from 2.5 to 18 MHz at HK and GZ. A straight line is drawn when the normalized spread-F occurrence rate is equal to 0% as in Fig. 12. The intersection of this line and the blue line is considered the foF2 threshold. We estimated that the foF2 threshold at HK and GZ was ~ 15 and ~ 14 MHz because almost the FSF occurrence was ~ 0% when foF2 exceeded this magnitude. The foF2 values ranged from of 3–9 MHz at BJ and CC. Thus, the corresponding foF2 thresholds for BJ and CC were 7.6 and 7.8 MHz, respectively. It is evident that the foF2 variability was much larger at low latitudes than at mid-latitudes. There are few reports that consider the effect of the foF2 threshold on the generation of spread-F events. De Abreu et al. (2017) found that the spread-F was often observed during storms using measurements from multiple instruments over the American sector when the foF2 was below 8 MHz. De Abreu et al. (2017) showed that the post-sunset EIA is produced by the plasma fountain arising from the pre-reversal vertical drift enhancement in the F-region (as indicated by large sunset increases of h′F and decreases of foF2). Therefore, it can be seen that the rapidly changing Dst index will also affect spread-Fs; however, our research is not currently focused on ionospheric storms. The variation in foF2 at different latitudes suggests that the PRE is not the only factor to initiate FSFs. For example, the meridional wind can suppress the growth rate of the R–T instability, also attributing to the foF2 and FSF (Buonsanto and Titheridge 1987; Stoneback et al. 2011).

Fig. 12
figure 12

Correlation between foF2 and the normalized FSF occurrence percentages

Full size image
Fig. 13
figure 13

Correlation between h′F and the normalized RSF occurrence percentages

Full size image

Figure 13 shows the post-sunset h′F variations compared with the RSF occurrence rates at the four sites. The red point is the sample value. The blue line is the fit curve. The RSF occurrence rate and the h′F satisfy the parabolic relationship. When the probability of the RSF was ~ 25% of the maximum probability of occurrence, we treated that virtual height value as the threshold value. The h′F occurring between 240 and 290 km is more favorable for RSF occurrence by calculation, which is different from the relationship between foF2 and FSF. Figures 6, 10, and 13 indicate that the higher occurrence rates of RSFs are well correlated with higher post-sunset h′F peaks (Rungraengwajiake et al. 2013). Previous studies observed spread-Fs in the equatorial region on nights when the h′F was below 300 km (Abadi et al. 2015; Manju and Madhav Haridas 2015; Liu and Shen, 2017; de Abreu et al. 2017). Our results also support this conclusion. In addition, Devasia et al. (2002), Jyoti et al. (2004) and Manju et al. (2007) obtained an h′F threshold for the spread-F occurrences in their studies in India. Devasia et al. (2002) found a threshold of about ~ 300 km for the cases in their study. Our results also show that when the virtual height is greater than 300 km, the probability of an RSF is very small. Jyoti et al. (2004) showed a linear relationship between solar activity and the h′F threshold. Manju et al. (2007) investigated the dependence of the h′F threshold on seasonal and solar activity for magnetically quiet conditions and proposed the important role of neutral dynamics in controlling the day-to-day ESF variability. Abadi et al. (2015) found that latitudinal extension of plasma bubbles was mainly controlled by the h′F peak value during the initial phase of an ESF. Manju and Madhav Haridas (2015) showed that the equinoctial asymmetry of the h′Fc increases with solar activity. In this article, the correlation between the h′F threshold and the seasonal and solar activities are not involved, and we will also focus on this content. The new idea presented from our study is the correlation between RSF occurrences and the h′F, which are different from previous research results. In a follow-up study, we will examine the relationship between the h′F threshold for RSFs and the solar and geomagnetic activities and equinoctial asymmetry.

The correlation RSF occurrence percentages with rapidly increasing post-sunset monthly mean h′F values substantiated the role of the PRE enhancement on RSF onsets. Traveling planetary wave ionospheric disturbance (TPWID)-type oscillations (de Abreu et al. 2014a, c; Fagundes et al. 2009) in the modulation of the virtual height in the F-region increased during sunset hours. Meridional wind velocities corresponding to the post-sunset h′F for each spread-F event have been considered. Buonsanto and Titheridge (1987) found that the hmF2 dropped from 13:00 to 18:00 LT during the solar maximum periods because of the meridional wind. These results also indicate that the spread-F is a complex phenomenon, which implies that other possible factors can be ascribed to spread-F occurrences. The atmosphere ionosphere coupling process has been proposed as a contributing factor for spread-F development. Therefore, the connections between spread-F occurrence characteristics and the foF2 and h′F magnitudes deserve detailed investigation by additional theoretical and observational research. The foF2 and h′F thresholds also require further investigation using observations from different regions and under different solar activity conditions.

Summary and conclusions

In this study, we presented variations of the spread-F, foF2, h′F, the possible threshold of the foF2 for FSF, and the relationship between the h′F and RSF. The data in our study were recorded by four stations at low- and mid-latitudes near 120°E longitude in China during the 23rd solar cycle. The major conclusions are summarized as follows:

  1. 1.

    The FSF occurrence rates increased during years of LSA at all four sites. FSFs mainly occurred during the summer months, while RSFs occurred mostly in the equinoctial months between 2000 and 2002 at HK and GZ. Post-midnight FSFs were the most observed type of spread-F events. The typical FSF onset time was about 21:00 LT, and the FSFs normally lasted until 05:00 LT, while the RSFs occurred 2–3 h earlier at HK and GZ during periods of HSA.

  2. 2.

    The foF2 and h′F peak values come mainly before midnight at low latitudes, while h′F peak values appeared after midnight at mid-latitudes during periods of HSA.

  3. 3.

    Lower foF2 values were appropriate for FSF events; nevertheless, h′F and RSF occurrences satisfied the parabolic relationship. Most FSF events occurred when the foF2 was below 15 and 14 MHz at HK and GZ, and below 7.6 and 7.8 MHz at BJ and CC. The h′Fs occurring between 240 and 290 km were more favorable for RSF occurrences, which differ from the foF2. However, some questions remain unresolved and further studies are in progress.

Our studies of FSFs and RSFs in China are useful and have the potential to be included in the future IRI model. However, even after such studies of spread-F onsets and growth conditions, some uncertainties remain. This requires further efforts to understand the spread-F phenomenon at different locations. Soon, long irregularity data coverage over the China sector will be studied. More ionospheric parameters will be compared with local time and seasonal spread-F variations to amplify knowledge of the involved physical mechanisms.