Recent Progresses on Ionospheric Climatology Investigations

The ionosphere varies over multiple time scales, which are classified into two categories: the climatology and weather variations. In this national report, we give a brief summary of recent progresses in ionospheric climatology with focus on (1) the seasonal variations, (2) solar cycle effects, and (3) empirical modeling of the ionosphere. The seasonal variations of the ionosphere have been explored in many works to give a more detailed picture with regional and global features at various altitudes by analyzing the observation data from various sources and models. Moreover, a series of studies reported the response of the ionosphere to solar cycle variations, which revealed some novel and detailed features of solar activity dependence of ionospheric parameters at different altitudes. These investigations have improved our understanding on the states of the ionosphere and underlying fundamental processes, provided clues to future studies on ionospheric weather, and guided ionospheric modeling, forecasting and related applications.


Introduction
The ionosphere is formed under the ionization effect of solar X-ray and Extreme Ultraviolet (EUV) radiation.At the same time, the ionosphere is buffeted by sources from solar winds and magnetospheric activities.The lower atmospheres also contribute considerably to the variability of the ionosphere.As a consequence, a variety of periodic and aperiodic variations are observed in the ionosphere, with time scales ranging from the order of the solar cycle (11-year) to just a few seconds.In this national report to COSPAR, the ionospheric climatology is termed the averaged be-haviors of the ionosphere with time coverage longer than the solar rotation period.
With the development of sounding techniques, the spatial coverage of ionospheric observations has got much better.Moreover, several global data centers have been established and opened for access, which ensure the data easily available for researchers.All of the above conditions make it possible for huge progress in ionosphere researches, particularly in climatology.
The seasonal variation, solar activity modulation, and long term trend of the ionosphere are fundamental problems of the ionospheric climatology.The studies on those issues are helpful for understanding ionospheric chemical and dynamical processes and have critical applications for empirical ionospheric modeling as well.
In this report, we give a brief introduction to the progresses of ionospheric climatology in recent several years.Our interest is mainly on (1) the seasonal variations, (2) solar cycle effects, and (3) empirical models of the ionosphere.The seasonal variations of the ionosphere have been explored to give a detailed picture with regional and global features at various altitudes by analyzing the observation data from various sources and models.Moreover, a series of works reported the response of the ionosphere to solar cycle variations, revealing novel and detailed features of solar activity dependence of ionospheric parameters at varying altitudes.These investigations have improved our understanding on the states of the ionosphere and underlying fundamental processes, provided clues to future studies on ionospheric weather, and guided ionospheric modeling, forecasting and related applications.

Seasonal Variation
The seasonal variation of the ionosphere is a traditional issue in ionospheric physics.It has been studied for decades, but there remain many unresolved problems.Chapman's ionization theory is usually used to predict the behaviors of the ionosphere.Sometimes, however, the peak density of F 2 -layer ionosphere (N m F 2 ) actually changes in a way inconsistent with the prediction of Chapman's ionization theory.Anomalies are taken to term the ionosphere under such situations with features significantly away from the theory's predictions.There are some anomalies in seasonal variations [1] .The principal anomalies observed include winter or seasonal anomaly (greater daytime N m F 2 in winter than in summer), semiannual anomaly (higher N m F 2 in equinoxes than in solstices), and annual anomaly or December anomaly (higher N m F 2 in December than in June).
To show the seasonal feature of the ionosphere, ionospheric key parameters are often analyzed to obtain the principal components with periods no longer than a year.Yu et al. [2] collected the critical frequency of F 2 -layer (f 0 F 2 ) recorded at global 104 stations to reveal the latitudinal dependence of annual and semiannual components of daytime N m F 2 for solar maxima.Their results show the following.(1)   The annual component of N m F 2 in both hemispheres dominates in the regions from 40 • to 60 • magnetic latitudes and becomes less significant at lower latitudes.Winter anomaly usually manifests strongly at higher latitudes and becomes less obvious or disappears at lower latitudes.(2) The semi-annual variations with peaks in equinox are generally weak in the near-pole regions and strong in the far-pole regions of both hemispheres.(3) In tropical regions the semiannual components are more significant than the annual components.
Yu et al. [3]  Through introducing the Empirical Orthogonal Function (EOF) analysis, Zhao et al. [4]  Liu et al. [5] collected the JPL GIMs to detect the overall climatology of TEC during 1998−2008 in a new way.They evaluated the mean TEC which were averaged over the globe, and over low, middle and high latitude bands, respectively.These mean TECs present that the annual and semiannual components are dominant in seasonal variations.The annual asymmetry between June and December can be detected easily in the global and low latitude mean TEC for all solar activities and in middle latitude and high latitude mean TEC under most solar activities.The annual variations of the mean TEC are more significant in the southern hemisphere, while the semiannual variations are in phase and have comparable amplitudes in both hemispheres.The interhemispheric asymmetry of the mean TEC averaged in one hemisphere follows the variation of solar declination.Both the inter-hemispheric and annual asymmetries become stronger with increasing solar activity.Afraimovich et al. [6] and She et al. [7] also conducted an evaluation of mean TEC, which were averaged over the globe.However, when we try to adopt the mean TEC for indicating the variability of the ionosphere, these TECs averaged over the globe and in both hemispheres as well should be used with caution.It may provide false information on the ionosphere, especial under low solar activity.Under some situations the ionosphere actually dominates with annual variations, while the mean TEC averaged over both hemispheres definitely show significant semiannual variations.
Zhao et al. [8]  Liu et al. [9] further presented some salient features in the 10-year (1996−2005) topside plasma density recorded by the DMSP in the magnetic latitude range of 60 • S−60 • N. Strong yearly variations are shown in the longitude-average topside plasma density.In particular, annual components are predominant at most latitudes with maxima around the June solstice in the Northern Hemisphere and the December solstice in the Southern Hemisphere.Seasonal anomaly only occurs in the northern equatorial regions.Moreover, the differences of plasma densities between the two solstices are not symmetrical about the magnetic equator, being generally higher in the Southern Hemisphere than in the Northern Hemisphere.Conjugate-average plasma density is substantially greater at the December solstice than at the June solstice, which is modulated by solar activity levels and has latitudinal and longitudinal structures.The longitude effects of the annual asymmetry depend on local time, being stronger in the evening sector than in the morning sector.Accumulated evidence indicates that the winter anomaly only occurs in a limited altitude range in the ionosphere.What we do not know is where the winter anomaly will disappear.Clues may be inferred from synthetical analysis of the observations from ROC-SAT (at about 600 km altitude), DMSP (840 km) and DEMETER (710 km) [10] satellites and those from COSMIC radio occultations as well.

COSMIC (a Constellation Observing System for
Meteorology, Ionosphere and Climate mission of six microsatellites) was successfully launched to 512 km altitude in April 2006.The six satellites initially were close to each other, then they moved gradually to orbits at around 800 km altitudes.The COSMIC radio occultation observations have accumulated more than 2 000 000 electron density (N e ) profiles, providing us an excellent 3-D monitoring of the ionosphere.Liu et al. [11] collected the COSMIC measurements to investigate the seasonal behaviors of daytime N e in the altitude range from 200 to 560 km.Harmonic analysis of N e at different altitudes provides unprecedented details of the seasonal behaviors of N e at Low Solar Activity (LSA).Fig. 1, taken from Fig. 1 of Liu et al. [11] , shows maps of A 12 (the annual amplitude), A 6 (the semiannual amplitude), Φ 12 (the annual phase), Φ 6 (the semiannual phase), A 0 (annual mean), A 12 /A 0 , A 6 /A 0 and A 12 /A 6 of the COSMIC N e at altitudes of 204, 250, 340 and 500 km at 13:00 LT.In each panel the white curve indicates the dip equator, and two magenta circles represent the locations of magnetic poles.A 12 , A 6 and A 0 are in units of 10 5 cm −3 , and Φ 12 and Φ 6 are in months.Furthermore, in equatorial regions, a new finding is the obvious wave-like pattern in the longitudinal structure of the amplitudes of seasonal harmonic components in equatorial regions, which supports possible couplings between lower atmospheric origins and the longitudinal variations of N e .Liu et al. [11] further illustrated the longitudinal structures of the key parameters of the F 2 peak (peak density and height and Chapman scale height) derived from the COS-MIC data.
The results presented by Liu et al. [11] indicated that the phase of the annual variation of h m F 2 is closely related to the sign of geomagnetic dip angle.
The annual component of h m F 2 definitely predominates in most regions, and the semiannual amplitude of h m F 2 is particularly weaker.Zhang et al. [12] also gave a picture of the seasonal variations of ionospheric profiles at high and middle latitudes, in terms of the measurements of incoherent scatter radars during the International Polar Year (IPY).A12/A0, A6/A0, and A12/A6 of the COSMIC Ne at 13:00 LT.In each panel the white curve indicates the dip equator, and two magenta circles represent the locations of magnetic poles.A12, A6 and A0 are in units of 10 5 cm −3 , Φ12 and Φ6 are in months.After Liu et al. [11] wever, it has not received enough attention in past investigations.For example, almost all ionospheric empirical models cannot reproduce the differences between March Equinox and September Equinox.In other words, these models did not treat the data in the two equinoxes separately.In the literature, there is not much information about the global picture for the equinoctial asymmetry.Balan et al. [13] collected the historical data observed from the Japanese MU incoherent scatter radar and showed higher electron densities in September equinox than in vernal months at lower altitudes.In contrast, the pattern is reversed at altitudes around and above the F 2 peak.They put forward that the equinoctial asymmetry in neutral wind should be responsible for the observed pattern in the topside ionosphere.

The equinoctial asymmetry is an interesting and open question in ionospheric seasonal variations. Ho-
Liu et al. [14] utilized multiply ionospheric data, including global measurements of N e profiles from COSMIC mission radio occultation measurements, ionosonde N m F 2 , TEC from TOPEX and Jason-1 and from global dual-frequency GPS receivers as well, to investigate the behaviors of the daytime ionosphere around equinoxes during Low Solar Activity (LSA).
The analysis reveals that during LSA the equinoctial asymmetry in ionospheric plasma density is mainly a low latitude phenomenon.The equinoctial differences in TEC and N m F 2 are remarkable at low latitudes in both hemispheres and less significant at higher latitudes.At higher altitudes, the asymmetry in COS-MIC N e becomes less significant in the southern hemisphere, and the pronounced asymmetry regions in the northern hemisphere move toward the magnetic equator.However, Chen et al. [15] recently revealed that the latitudinal features of the equinoctial asymmetry of N m F 2 and TEC at high solar activity are different from those at low solar activity.At high solar activity, the equinoctial asymmetry of N m F 2 is more prominent at middle latitudes while the equinoctial asymmetry of TEC is significant at all latitudes.In fact, the ionospheric equinoctial asymmetry should be considered as a manifestation of the annual variation, whose annual phase significantly shifts away from the solstices.According to this investigation, future models should treat the ionosphere on the two equinoxes separately, at least in the low latitude regions.
The F 2 -layer peak height (h m F 2 ) estimated from the COSMIC N e profiles also shows an equinoctial asymmetry at low latitudes, indicating the existence of equinoctial differences in low latitude neutral winds, specifically in the northern hemisphere.However, the pattern of equinoctial asymmetry in h m F 2 does not support the explanation of the absolute role of neutral wind effect on that of N m F 2 as proposed by Balan et al [13] .Therefore, besides the important effect of the neutral wind, there should be other physical processes playing key roles in forming the observed equinoctial asymmetry in the ionosphere.
Ren et al. [16] investigated the equinoctial asymmetry in the vertical plasma drift velocity in equatorial F regions based on the ROCSAT-1 observations during the period from 1999 to 2004.The ROCSAT-1 ionospheric observations exhibit noticeable asymmetry in plasma drifts with obvious local time dependency.On the basis of the plasma drift observations, they also simulated the effect of the equinoctial asymmetry in plasma drift on ionospheric plasma density with the Theoretical Ionospheric Model of the Earth, Institute of Geology and Geophysics, Chinese Academy of Sciences (TIME-IGGCAS).The simulations indicate that the daytime equinoctial asymmetry in plasma drift could partly explain that in plasma density observed by Liu et al [14] .
Currently, the atmospheric circulation mechanism is a widely accepted hypothesis for the ionospheric seasonal variations [1] .Solar radiation and Joule heating in the polar region are the main heating sources for the upper atmosphere.The input energy controls the convection pattern of the atmospheric circulation, changes the atmospheric composition, and further modulates the seasonal variation of the ionosphere.A simplified sketch is the summer-towinter flow, which is driven by solar heating.In the summer hemisphere rich-N 2 is carried upwelling with the heated air, and downwelling occurs in the winter hemisphere; as a result, the [O]/[N 2 ] ratio decreases in the summer hemisphere and increases in the winter hemisphere.On the other hand, the summer-towinter flow is reinforced by the effect of the summer auroral oval.Joule heating in the winter hemisphere also drives a subsidiary circulation [1] .Fuller-Rowell et al. [17] proposed the effect of "thermospheric spoon" to explain the semiannual variation, which is significant in the thermosphere.
According to their theory, the cross-hemispheric general circulation acts as a spoon in the summer hemisphere to better mix the thermospheric compositions, compared to equinoctial months.The consequence is that the thermosphere as a whole is more molecular (i.e., lower [O]/[N 2 ]) at solstice than at equinox.This also explains the F 2 -layer semiannual effect in N m F 2 prevailing at low and middle latitudes.
Oliver et al. [18] discussed the physical causes of middle latitude F layer behaviors.They emphasized the effect of neutral winds on h m F 2 and the composition control of N m F 2 .Ma et al. [19−20] studied the semiannual variations in low-latitude f 0 F 2 and in equatorial electrojet and proposed that the lower thermospheric diurnal tides modulate the semiannual variation of equatorial electric fields, which causes the semiannual variation in the low latitude ionosphere.
Yu et al. [21]  Another open question is the annual asymmetry in F 2 layer.Although this feature has been detected for decades, its principal causes are not fully understood.Zeng et al. [22] [22] declared that the tides originating from lower atmosphere only have a minor contribution to this asymmetry, its quantitative contribution is not clear up to now.Accumulated evidence indicates the tides originating from the lower atmosphere closely control the temporal and spatial structure of the ionosphere [23] .
For example, the longitudinal ionospheric structure of wave number 4 may be linked with the atmospheric tide of nonmigrating tidal mode DE3 (diurnal eastward wave number 3).It is surely a challenge topic on the contributions of tides to the detailed structure of seasonal variations in the ionosphere.
Using ionosonde data from 33 stations in three longitude sectors from 1969 to 1986, Xu et al. [24] detected a periodic oscillation in daytime N m F 2 with a period of 4 months (terannual), which is stronger in the midlatitude region in the Northern Hemisphere than in the Southern Hemisphere.In contrast, the terannual oscillation signature is not obvious in the nighttime F 2 layer.The phases match perfectly between the annual and semiannual oscillations and the terannual oscillation.These three oscillations show larger amplitudes during solar maximum.The amplitude of the terannual oscillation is also correlated with the product of the amplitudes of annual and semiannual oscillations.According to the above features, it is suggested that the terannual oscillation might be related to the nonlinear interaction between the annual and semiannual oscillations [24] .

Solar Cycle Variations of the Ionosphere
It is well known that the ionization effect of solar radiation principally forms the Earth's ionosphere.
The intensity of solar EUV is found to fluctuate regularly and irregularly over timescales from minutes (flares), roughly 27 days (solar rotation) to decades (11-year solar cycle), with amplitudes varying up to more than 1000 times [25−26] .Accompanying the variations in solar activity, many parameters of the ionosphere/thermosphere fluctuate with similar periods and features [27−29] .Therefore, the solar activity effects on the ionosphere are essential and complicated.
Under the invitation from the editorial committee of Chinese Science Bulletin, Liu et al. [29] summarized some recent studies on solar activity effects of the ionosphere.Their short review reported some observations of solar irradiance at X-ray and EUV wavelengths and the outstanding problems of solar proxies in the view of ionospheric studies, new findings and improved representations of the features of the solar activity dependence of ionospheric key parameters and the corresponding physical processes, possible phenomena in the ionosphere under extremely high and low solar activity conditions, and statistical studies and model simulations of the ionospheric response to solar flares.In the following, some materials are taken from this review to keep the completeness of this national report.The reader is recommended to refer to Liu et al. [29] for details.

Solar Cycle Effects of the F 2 Peak
The F 2 peak is the most important region in the ionosphere, due to the highest electron density around the F 2 peak.Therefore, it has been the subject of most ionospheric investigations.Recent works studied some key parameters, such as f 0 F 2 or N m F 2 [30−37] , TEC [6−7,38−39] , h m F 2 [31,34] , plasma temperature and profile scale heights [31−40] , thermospheric winds [41−43] , and total mass density of the thermosphere [44] .These studies investigated the complicated responses of the ionosphere and thermo-sphere to changes in the solar activity, the related mechanisms, and how to represent those characteristics effectively in application missions [15,34−35] .
The picture of solar activity effects of the ionosphere is gradually outlined in a more detailed way [30,32−33,46−47] .Many investigations have been conducted on the statistical relationship between ionospheric electron density and the solar proxy F 10.7 .
For example, Balan et al. [46] , Liu et al. [32] and many others reported that the increase in TEC and N m F 2 linearly follows the growth of F 10.7 under low and moderate solar activity levels, but the linearity breaks down at higher activity levels.Under such situations, TEC and N m F 2 stop to increase and even turn to decrease at some stations.
It is under debate the nonlinear feature of the solar effects of the ionosphere and what causes such nonlinearities.Balan et al. [46−47] originally explained the saturation in electron density in response to the solar activity in terms of nonlinearity between solar indices and EUV intensity.Traditionally, the sunspot number R and F 10.7 are used for solar proxies.
However, Liu et al. [32] reported that the saturation effect is also detectable in f 0 F 2 and EUV, and becomes stronger in equatorial anomaly regions.
Liu et al. [34] further carried out an analysis based on a longest series of daily solar EUV data recorded by SOHO/SEM and N m F 2 data recorded at global ionosonde stations in the East Asia/Australia sector.
Their original objective was to quantify the solar activity sensitivity of daytime N m F 2 .Liu et al. [34] found that dynamic and chemical processes determine the seasonal and latitudinal features of the saturation effect.According to their results, the nonlinearity between solar EUV and F 10.7 is not enough to explain the observed saturation effect.
Kane [48] puzzled the inconsistency between the change rates in solar EUV intensity and those in N m F 2 .For example, solar EUV increased 150% from 1996 to 2000, while N m F 2 changed from 210% to 290% at seven stations.Obviously, the solar EUV changes themselves were insufficient to account for the observed changes in N m F 2 .It means that the solar activity not only determines the photoionization rates but also modulates various related processes.Liu et al. [34] reproduced the observed inconsistency of the change rates of solar EUV inten- Chen and Liu [49] further studied the influence A novel finding is that the solar activity effects of ionospheric electron density show linear, saturation and amplification features.The trends the ionosphere presented are found to be dependent on latitude, local time and season [5,30,39,50−51] .The solar activity variation in nighttime N m F 2 at some stations retains the daytime saturation feature in summer, has a linear feature during the equinoxes, and presents an amplification feature in winter.Chen et al. [30] stated that dynamical processes (pre-sunset enhancement of the equatorial E×B vertical drift and neutral winds) and changes in neutral compositions contribute to the seasonal difference in the solar activity effects of nighttime N m F 2 .Liu and Chen [39] provided a quantitative description of the three kinds of solar activity effects in the global TEC.They illustrated the local time variation and longitudinal distribution of the TEC solar activity sensitivity in the four seasons.The saturation is found to cluster in equatorial anomaly regions, which were supported by Ma et al [52] .Ref. [52]   used ionosonde N m F 2 data to show latitudinal double peaks in the nonlinear coefficient of N m F 2 versus solar proxies.
Recently, there are some attempts to search for usable ionospheric indices.Afraimovich et al. [6] , She et al. [7] and Liu et al. [5] evaluated the mean TEC and its solar signature.These mean TEC data effectively depress the noise from local regions and capture obviously various time-scale variations in the solar activity [25,48,53] , compared with individual TEC data.Moreover, the mean TEC averaged over latitude bands shows stronger solar activity sensitivity in lower latitude bands, and the saturation effect in the mean TEC versus F 10.7 is more pronounced at low latitudes, while the mean TEC increases more rapidly for higher solar EUV fluxes [5] .
A long-lasting question in solar cycle effects of the ionosphere is the "hysteresis" phenomenon [38] .
Similar to the "hysteresis" effect of magnetic materials, electron density may have different values for different phases of a solar cycle even at the same solar level.There is not yet an accepted explanation for the "hysteresis" effect.Attempts have been made in ionospheric models to include the possible influence of the historical state of the solar activity on the ionosphere [3,33] .The "hysteresis" effect, however, is difficult to reproduce effectively in nowadays models, which is one of the key problems restricting the development of long-term predictions of the ionosphere.

Solar Cycle Effects of the Ionosphere at Different Altitudes
To first order, the ionosphere varies with altitude.The solar cycle effect on ionospheric N e profile and ion compositions is an interesting issue.Zhao et al. [18] studied the changes in topside ion compo- Under the assumption of diffusive equilibrium, the altitudinal distribution of the topside plasma density is determined by h m F 2 , N m F 2 and plasma scale height.The plasma scale height naturally is an important parameter to describe the altitudinal distribution of plasma density in the topside ionosphere.In the light of this simplification, the peak parameters and plasma scale height can be derived from the observations of incoherent scatter radar.Lei et al. [31] and Liu et al. [40] systematically investigated the climatology of plasma scale height around the Liu et al. [9,51] presented the climatology of plasma density at 840 km altitude.The data are the total ion density observed by DMSP.It is the first time to report that there are strong amplification features in the solar activity dependency of plasma density in the topside ionosphere.This feature becomes more obvious at lower magnetic latitudes.What is surprising to us is that the solar activity dependency of electron density becomes much complex, when the altitude moves downward to 600 km altitude [50] .
ROCSAT-1 observations indicate that all the three patterns, linearity, saturation and amplification, can be found at 600 km altitude.In contrast, the saturation effect predominates at 400 km altitude [54] .Moreover, the equatorial anomaly feature is fairly obvious in the latitudinal distribution of ROCSAT-1 sunset plasma density, indicating the enhancement in equa-torial vertical plasma drift during solar maximum.
According to the characteristics of ionospheric Flayer key parameters, Liu and Chen [17] , Liu et al. [51] and Chen et al. [50] proposed an explanation to understand why there are three kinds of patterns at different altitudes in the solar activity response of the ionosphere.The principal factor controlling the solar activity variation trend of the plasma density at high altitudes (e.g., 800 km) is the scale height, which drives the increase of the ionospheric plasma density in an exponential way.With the altitude moves downward, the scale height, h m F 2 and N m F 2 all play key roles.In particular, equatorial drift induces the change in h m F 2 , which further controls the solar activity response of the equatorial ionosphere.

Solar Conditions
The solar activity evolution itself is a very complicated and elusive process, and some extreme phenomena have been reported emerged in historical solar activities.For example, according to sunspot records on hand, virtually no sunspots were observed during the period of 1645−1715 (the Maunder Minimum), which indicates the Sun experienced an extremely long and low solar activity period.
Liu et al. [5] investigated the statistical relationship between ionospheric TEC and solar EUV.Negative values of plasma density are extrapolated, if one applies the statistical function of observed TEC and EUV to the situation that the Sun ceases to radiate EUV outward.The negative extrapolated TEC is certainly a non-physical result because the value of TEC should not be negative.However, it reveals a key point that under extreme solar conditions the state of the ionosphere should be different significantly from those under normal conditions.This gives rise to exploring how the ionosphere performs under such extremely low solar EUV radiation.
Smithtro and Sojka [55]  as those of N m F 1 , similar to the "G" condition in some storms.One feature we do not expect is that the simulated electron density grows faster when the solar activity goes beyond its normal range, which is consistent with that of the mean TEC observed in Liu et al [5] .
It should be mentioned that the solar activity in Chen et al. [45] found that the ability of Solomon et al. [56] reported the total mass density of the upper atmosphere became smaller based on the atmospheric mass density database retrieved from the historical data of orbits of thousands of satellites.
The total mass density normalized at 400 km altitude fell below 30% of those during previous solar minima, which means that the atmosphere became cooler.
Lühr and Xiong [57] compared the observed electron density from CHAMP and GRACE satellites with the predictions of an empirical ionospheric model [58] and showed that the prediction of the empirical model overestimates the ionospheric electron density in 2007−2009.In some sense, it also indicates the ionosphere behaviors were unusual during this period.Heelis et al. [59] also reported the ion temperature and the O + /H + transition height became lower during this period.
Recently, Liu et al. [60] analyzed the global data of ionosondes and TEC GIMs during solar minima.10 < Ap 20, 5 < Ap 10, and Ap 5, respectively.After Chen et al. [45] minimum in 2009.We can infer from the above studies that during the deep solar minimum the ionosphere responses in lower height and smaller electron density may be mainly attributed to the decrease in solar EUV.
As illustrated in Fig. 3, the 1-year running mean value of f 0 F 2 drops 0.55 MHz at a Japanese station and the new proxy F 10.7P decreases about 3 sfu (solar flux unit).The work of Liu et al. [61] using COS-MIC radio occultation measurements also supports this trend.Decrease in the intensity of solar EUV is proposed to act as the main driver for this decrease in ionospheric electron density.It deserves to stress that a quadratic fitting reasonably captures the solar variability of f 0 F 2 and global average TEC at such low solar activity levels.
Araujo-Pradere et al. [62]  Unfortunately, we still lack the knowledge of solar EUV spectrum under extreme solar activity conditions from observations.Therefore, we have difficulty to preciously reproduce the picture of the ionosphere under such situations.

Ionospheric Empirical Modeling
Ionospheric sounding operated at Wuhan Ionospheric Observatory (WIO, 114.4 • E, 30.6 • N) has a history over 60 years, so WIO belongs to one of the stations with longest routine ionospheric observations.A series of single station models have been constructed on the basis of WIO observations.For example, Chen et al. [63] and Mao et al. [64] developed empirical models of TEC using the Fourier expansion and EOF repre- Here, F10.7A is the 81-day centered mean of F10.7.The minimum value of f0F2 during solar minimum of cycle 23/24 is at least 0.553 MHz lower than those of the preceding cycles, and that of F10.7P in cycle 23/24 is 3.06 sfu lower than those in the preceding cycles.After Liu et al. [60] sentation methods, respectively.Liu et al. [33] built models of f 0 F 2 using Fourier expansion and cubic-B splines approaches.The f 0 F 2 models have good prediction accuracies with standard deviations in about 0.26 to 0.58 MHz.Yue et al. [65] gave a model regarding Wuhan f 0 E. Liu et al. [66] adopted EOF method to develop a model of M(3000)F 2 .In addition, through the work by Zhang and Huang [67] , the bottom-side observations over Wuhan have been introduced in the International Reference Ionosphere (IRI), which has already improved the model description for the ionosphere in China.
Besides the data over Wuhan, Xu et al. [37] reconstructed a single station model using f 0 F 2 data at Chongqing (29.4 • N, 106.5 • E).
Recently, Wan et al. [68] and E. A. et al. [69] con- On the basis of the EOF base functions, Wan et al. [70] developed a now-casting system of TEC maps over China with real-time observation of TEC from four stations, Mohe, Beijing, Wuhan and Sanya, which distribute from the north to south of China.
In another work, E. A. et al. [71] discussed the impact of different model drivers through choosing appropriate proxies to present f 0 F 2 .They used similarparameter interpolation method to optimally refill the missing points in original f 0 F 2 series and applied an EOF analysis to construct empirical f 0 F 2 models.
The data are from three Japanese ionosonde stations, Wakkanai (45.4The diurnal variation of base functions reflects the essential nature of ionospheric f 0 F 2 , and the coefficients represent the variability.For example, the first order EOF coefficient reflects the solar cycle variation in the component.
Zhang and Holt [72] established an ionospheric electron temperature model based on incoherent scatter radar observations at Millstone Hill of America.
Zhang et al. [73] built an ionospheric plasma temperature (ion and electron temperatures) model over French Saint Santin with a slightly higher geographic latitude but an lower apex latitude (14 Liu et al. [74] adopted the autocorrelation method to forecast the short-term variations in f 0 F 2 .They estimated the prediction errors to examine the forecast performances at Chongqing for different combinations of parameters and algorithms.Their work suggests the "at once" method for longer than 10 h predictions and the "iterations" method for shorter predictions.At the same time, the CRI model, a corrected method of IRI, was described.An effective ionospheric index was introduced into the CRI model to realize the regional forecasting in China region. Yue et al. [75] made the first attempt to use an Artificial Neural Network (ANN) method to derive the possible long-term trend in f 0 F 2 .They analyzed f 0 F 2 data at 19 ionospheric stations in the Asia/Pacific sector.A byproduct of their analysis is that the ANN method can effectively eliminate the geomagnetic activity effect on f 0 F 2 , compared to usual regression methods.
M(3000)F 2 is a key parameter to estimate h m F 2 .
Comparisons of empirical models with observations show a huge challenge for improving the presentation of M(3000)F 2 .The CCIR option in IRI provides M(3000)F 2 with significant differences from observations, particularly in low latitude and equatorial regions.
Zhang et al. [76−77]  to give some suggestions to future modeling for the ionosphere: (1) recommend F 10.7P [34] as solar proxy instead of F 10.7 and sunspot number; (2) adopt a quadratic approximation to describe the solar variability of electron density, instead of a linear fitting or other higher-order regression fittings; (3) include the equinoctial differences of the ionosphere; (4) effectively decrease the terms of the base functions in models with new analysis approaches (e.g., EOF).
gave a global distribution of the annual and semiannual components of Total Electron Content (TEC) in 2000 (a year at solar maximum) through retrieving the TEC data from the international Global Positioning System (GPS) service center (IGS).Their results show that the annual amplitudes of daytime TEC have larger values at middle and high latitudes in both hemispheres and smaller in equatorial and low latitude regions.In contrast, the semiannual amplitudes are larger in far-pole regions than in near-pole regions.Moreover, in most regions the peak TEC appears in equinoctial months, while the maximum values appear in winter in northern near-pole regions and in summer in South America and Australian.Similar to N m F 2 , the seasonal variation of TEC is mainly attributed to the combined effects of changes in atmospheric composition [O]/[N 2 ] and the photo-ionization production rates controlled by solar zenith.
determined the EOF spatial base functions and temporal base functions in daily evolution of global distribution of the Jet Propulsion Laboratory (JPL) Global Ionospheric Maps (GIMs) of TEC during the period from 1999 to 2005 and further presented the seasonal variation of TEC.The EOF spatial bases revealed significant seasonal anomaly in the time interval of 10:00 LT−15:00 LT at middle latitudes and semiannual anomaly in equatorial ionization anomaly region being strongest near 20:00 LT and weakest around sunrise.The EOF temporal bases indicate strong solar cycle effects on the seasonal variations of TEC.The contributions of seasonal anomaly and annual anomaly to the variability of TEC are larger than those of semiannual variations.They also found prominent equinoctial asymmetries in East Asia and South Australian regions.Further analysis of TIMED GUVI columnar [O]/[N 2 ] data shows that the [O]/[N 2 ] variation is a major contributor to the daytime winter anomaly of TEC.
applied an EOF analysis to examine the climatology of the topside ionosphere.The data used are the total ion density N i at 840 km altitude, observed by the Scintillation Meter onboard the Defense Meteorological Satellite Program (DMSP) spacecraft in two Local Time (09:30 LT and 21:30 LT) sectors during the period of 1996−2004.The first EOF mode of the N i variability is modulated by the intensity of solar EUV flux.The second EOF presents an annual asymmetry.The semiannual variation appears in the third EOF mode for the evening sector.The N i variability is rapidly convergent; therefore it is very convenient to adopt the EOF expansion to construct an empirical model.The prediction accuracy of the model depends mainly on the first few principal components.

Fig. 1
Fig. 1 indicates that even during solar minimum there are strong and complicated annual and semiannual variations in daytime N e , which have distinct latitudinal and altitudinal dependencies.The semiannual component predominates over the annual variation in equatorial regions and at high latitudes in

Fig. 1
Fig. 1 Maps of A12 (the amplitude of the annual component), A6 (the amplitude of the semiannual component), Φ12 (the phase of the annual component), Φ6 (the phase of the semiannual component), A0 (annual mean), conducted simulations to examine the control of electric fields on N m F 2 under solar minimum and geomagnetic quiet conditions.Their results indicate that the semiannual variation of electric field would strengthen that of N m F 2 over equatorial anomaly regions in the afternoon to early morning hours.
investigated global variations of the annual asymmetry observed from 1year COSMIC ionospheric radio occultation measurements.The COSMIC observations reveal strong annual asymmetry in N m F 2 with significant longitudi-nal and local time variations.The numerical simulations using the Thermosphere-Ionosphere Electrodynamics Global Circulation Model (TIEGCM) are in very good agreement with the COSMIC observations.TIEGCM simulations indicate that differences in solar EUV radiation between the December and June solstices (due to the change in the Sun-Earth distance) and the displacement of the geomagnetic axis from the geographic axis are the two primary processes that cause the annual asymmetry and its associated longitudinal and local time variations.Although Zeng et al.

sity and N m F 2 ,
if the solar activity effects of neutral compositions and chemical and dynamical processes are consistently taken into account.They further recommended an improved solar proxy F 10.7P (here F 10.7P = (F 10.7 +F 10.7A )/2, F 10.7A is the 81-day mean F 10.7 ), because F 10.7 and sunspot number fail to linearly represent the intensity of solar EUV.
of ionospheric dynamics and thermospheric variations on solar cycle effects of daytime N m F 2 .It is found that the variations of neutral atmosphere and h m F 2 , which is modulated by dynamic processes, are responsible for the seasonal and local time dependences of the linear increase rate of N m F 2 on F 10.7 at midlatitudes, while dynamic processes are the more important factors at low latitudes.Under high solar activities, dynamical processes not only control h m F 2 to influence N m F 2 , but also directly slow down the change rate of N m F 2 versus F 10.7 , which is a key factor for causing the saturation effect.

F 2 layer
peak, which are estimated from the observations of incoherent scatter radars at Arecibo (18.5 • N, 66.8 • W) and Millstone Hill (42.6 • N, 288.5 • E).The scale height linearly increases with F 10.7 and its variations have close relations with the thermal structure and dynamical processes in the ionosphere.They also discussed the possible connection between the scale height and N e profile parameters, which has potential values in ionospheric models.Some parameters may be estimated from bottom-side profile parameters, providing the connection is strong enough.
conducted a pioneer work, in which they simulated the global average response of the ionosphere and thermosphere to different solar activities in terms of a global average ionospheric/thermospheric model with assumptions of the solar EUV spectrum under extreme conditions.Their simulated results show that the change in neutral temperature and h m F 2 with solar EUV intensity still keeps a linear rate even under extreme low solar conditions.With solar EUV decreasing to very low level, the molecular N 2 and O 2 are dense around the F 2 peak so that the electrons have a larger loss rate; as a consequence, the decrease in O + concentration is faster than that in molecular ions.Under such situations, the values of N m F 2 become as low enough 2007−2009 was low and extremely extended in recent solar minima.The spotless days reached 266 days in 2008, a record in 90 years.A campaign named "Deep Solar Minimum" is going on to investigate the response of the space environment to such unusual solar conditions.The behaviors of the ionosphere and thermosphere naturally become the hot spot in the international community of space physics.
F 10.7 to indicate the intensity of solar EUV became worse during the period of 2007−2009.It means that the so-lar activity is surely unusual.Geomagnetic activity was much more quiet during the recent deep solar minimum than in the previous several solar minima (Fig. 2).A consequence of the calm geomagnetic activities is that the separation of atmospheric compositions accelerates, and the ratio of [O]/[N 2 ] increases.
Fig. 2 Dot curve shows temporal series of F10.7 during recent solar cycles, and bars show percentage distributions of Ap during solar minima (F10.7 < 80 sfu).Four segments, from top to bottom, of each bar for Ap > 20, also detected lower TEC over 4 middle latitude GPS stations in 2007−2009.It is interesting that they found a minor and marginal variation in ionosonde N m F 2 .Therefore, different studies reported inconsistent changes in N m F 2 from solar minimum to minimum; which certainly calls for further investigation to gain a better understanding of the ionospheric behavior under such unusual solar conditions, to examine the data consistency, and to determine the influence of analysis techniques applied.It should be pointed out that Araujo-Pradere et al.'s analyses are based on the monthly median values in selected months, while others are 1-year running values.The conflict possibly reflects the different variability of the ionosphere over different timescales.
ducted a complete EOF analysis of JPL-TEC and further constructed a TEC model.The global TEC model is constructed using the 1999−2009 GIMs provided by JPL.Their analyses also showed that EOF has outstanding advantage of fast convergence: the first four orders of the EOF series represent 99.04% of the overall variance of the original data set.Modeldata comparisons indicate that the model can reflect the majority of the variations and the feature of temporal-spatial distribution of the global ionospheric TEC.
made attempts to develop global models of M(3000)F 2 and h m F 2 .They also based on EOF methods to present the global climatology of the monthly median hourly values of ionosonde M(3000)F 2 .The models were verified using data in 1965 (a low solar activity year) and 1970 (a high solar activity year).Better performance than IRI was showed by the comparisons between the model and observations.The empirical presentation of the temporal and spatial variations of the ionosphere has vital values in applications.It is a long-lasting task.Here we wish • ) than Millstone.The French Saint Santin incoherent scatter radar observations covered over two solar cycles beginning in September 1965.Both mid-latitude ionospheric temperature models provide the climatological thermal status of the upper atmosphere.Comparisons with IRI predications indicate good agreement in ion temperature at high solar activity, and above the F 2 peak, while IRI electron temperature tends to be higher than both the St. Santin and Millstone Hill models.