Near-earth solar wind flows and related geomagnetic activity during more than four solar cycles (1963-2011)

h past studies, we classified the near-Earth solar wind into three basic flow types based on inspection of solar wind plasma and magnetic field parameters in the OMNI database and additional data (e.g., geomagnetic indices, energetic particle, and cosmic ray observations). These flow types are: (1) High-speed streams associated with coronal holes at the Sun, (2) Slow, ioterstream solar wind, and (3) Transient flows originating with coronal mass ejections at the Sun, including interplanetary coronal mass ejections and the associated upstream shocks and post-shock regions. The solar wind classification in these previous studies commenced with observations in 1972. In the present study, as well as updating this classification to the end of 2011, we have extended the classification back to 1963, the beginning of near-Earth solar wind observations, thereby encompassing the complete solar cy cles 20 to 23 and the ascending phase of cycle 24. We discuss the cycle-to-cycle variations in near-Earth solar wind structures and l1e related geomagnetic activity over more than four solar cycles, updating some of the results of our earlier studies.


Introduction
In past studies (Richardson et aL 2000(Richardson et aL , 2001(Richardson et aL , 2002, we divided the near-Earth solar wind since 1972 into three basic flow types in order to assess, for example, the contribution of each type of solar wind flow to long-tenn (> .... olar rotation) averages of geomagnetic indices and the intetplanetary magnetic field, and to detennine the structures driving geomagnetic stonns. The three flow types are: • Corotating high-speed streams, typically with solar ~d speed v> -450 kIn s-', that originate in coronal holes at the Sun (Krieger et aL 1973 ;Zirker 1977). The properties of corotating high-speed streams near 1 AU were summarized by Belcher & Davis (1971) and include the formation of a region of compressed plasma, the "corotating interaction region" (CIR), at the leading edge of the stream as it interacts with the preceding slower, cooler, and denser solar wind. Since the source coronal holes may persist for longer than a solar rotation, a given stream may recur at the solar rotation period (-27 days as viewed from Earth).
• Slower, interstream solar wind, typically associated with the streamer belt at the Sun (e.g., Feldman et aL 1981 ); and • Transient flows originating with coronal mass ejections (CMEs) at the Sun, including intetplanetary coronal mass ejections (lCMEs), the manifestations in the solar wind of CMEs, and the associated upstream shocks and postshock/sheath regions (see Zurbuchen  and references therein for discussion of the in-situ Signatures ofiCMEs). We collectively term these "CME-associated" flows.
In recent years, solar wind composition/charge state observations have been used to make a similar classification of flows during solar cycle 23 (Zhao et al. 2009). Unfurtunately, such observations are not generally available for earlier cycles, so other data have to be used. In om studies, the solar wind flow classification is based on inspection of a variety of data These include I-hour averages of near-Earth solar wind parameters obtained from the OMNI2 (formally OMNI) database (http:// omni"eb.gsfc.nasa.govi; King & Papitashvili 2005). The OMNI2 data extend back to 27 November 1963 and are compiled from observations made by various near-Earth spacecraft that have been carefully intercalibrated. OMNI2 data coverage is variable, with significant gaps in the early data, as will be discussed further below, and from 1983 until late 1994, when solar wind observations were predominantly made by IMP 8, which only spent part of each orbit of the Earth outside the Earth's bow shock. The ICME identifications made in om previous studies (e.g., Richardson & Cane 1993;Richardson & Cane 1995;Richardson et al. 1997;Cane & Richardson 2003;Richardson & Cane 2010) have been incOtporated into the solar wind classification. However, because the data required to separate ICMEs from their upstream "sheaths" are not consistently available, in particular in near-Earth observations prior to solar cycle 23, we do not differentiate between these structures when classitying solar wind flows, referring to them collectively as "CME-associated" flows as noted above.
Geomagnetic activity data are also examined since activity may be enhanced during the passage of ICMEs and associated flows (e.g., Burlaga et aL 1981 ;Wilson 1987Wilson , 1998Tsurutani et al. 1988;Gosling et al. 1991 ;Tsurutani & Gonzalez 1997; Sheeley et a1. 1976Sheeley et a1. , 1977Richardson et a1. 2006;Tsurutani et a1. 2006, and references therein). Examples will be illustrated below. Increased geomagnetic aetivity is associated with enhaneements in the )"Component of the solar wind convective electric field E ~ -V x B, i.e., Ey -VB" where B, is the southward magnetic field component:, which leads to enhanced reconnection between solar wind and magnetospberic magnetic fields and enhanced energy deposition into the magnetosphere (e.g., Dungey 1961 ;Tsurutani & Gonzalez 1997;O'Brien & McPherrun 2000;Ji et al. 2010, and references therein). In the case of CME-""",ciated flows, the southward magnetic fields may be in the ICME and/or the upstream sheath formed between the ICME and associatlld shock (e.g., Tsurutani et a1. 1988;Huuunen& Koskinen 2004;Zhang et a1. 2007). In high-speed streams, intennittent intervals of southward fields associatlld with Alfveruc ftuctuations moving out fiom the Sun resuh in geomagnetic aetivity that may persist for several days during passage of a stream past the Earth, and recur at the solar rotation period (e.g., Sheeley et al. 1976: Burlaga & Lopping 1977Tsurutani & Gonzalez 1987;Tsurutani et a1. 2(06). Such activity can help to indicate the presence of streams when no solar wind speed observations are available. In addition, geomagnetic stonn sudden commencements (SSCs) can help to identifY the passage of interplanetary shocks at the Earth (Gold 1955), which may be generated ahead of fast ICMEs or CIRs, with the caveat that not all SSCs are caused by shocks (e.g., Gosling et al. 1967;Chao & Lepping 1974;Wang et al. 2006). Energetic (-D. 1-100 MeV) particle observations (principally made by Goddard Space Flight Center instruments on various near-Earth spacecraft) are also considered since these can help to indicate the passage of shocks and ICMEs. Solarparticle event intensity time profiles often peak around shock passage, especially at lower eneIgies, and then may fall abroptly as the ICME arrives a few hours after shock passage (e.g., SandelSOn et al. 1990;Cane & Lario 2006;Klecker et al. 2006, and references therein), Modulations in the galactic cosmic ray intensity (i.e., "Forbush decreases"; Forbush 1937) can help to identifY the passage of shocks and ICMEs (e.g., Barnden I 973a, 1973b;Cane et a1. 1993Cane et a1. , 1996Cane 2000;Richardson & Cane 2011b, and references therein) and also corotating streams (Simpson et a1. 1955;Iucci et a1. 1979;Richardson et a1. 1996;Simpson 1998;Richardson 2004, and references therein). We have used cosmic ray observations flum neutron monitors and also fiom spacecmft, in particular the coonting mte of the anti-coincidence gnard of the Goddanl instrument on [MP 8 (e.g., Cane 1993;Richardson et a1. [999). In summary, by combining these various data sets, we have been able to make a reasonably complete classification of the solar wind structures at Earth even when the solar wind data are not always complete.
Since these earlier studies, we have continued to update the solar wind How classification to near-present This has proved to be particularly valuable for studies of the magoetospheric and ionospheric response to differeot types of solar wind structures (e.g., Emery et a1. 2009Emery et a1. , 2011Turner et a1. 2009) and is available at the CEDAR workshop web site (bttp '/cedarweb.hao. ucar.edulwiki/index.php,Tools_and_Models:Solar_ Wind_ Structures). We bave also recently extended the solar wind classification fiom 1972 back to 1963. This will be discussed in the next section, and examples of obselVations from this period will be illustrated. We then discuss the solar cycle variation in solar wind parameters and structures from 1963 to 2011. Section 4 summarizes the results of the paper.
2. Extension of the solar wind classIfIcation back to 19631solar cycle 20 As noted above, solar wind observations incorpomted into the OMNI2 database are available back to 1963 but our previous studies did not consider the period before 1972, when the observations become more complete (at least until 1983). Since these earlier observations encompass sunspot cycle 20, the weakest cycle so far during the space era, and with cycle 24 also expected to be a relatively weak cycle (e.g., http:// www.swpc.noaa.goY·SolarCycl .. ·SC24Iindex.html). we have extended the solar wind flow classification back from 1972 to the beginning of in-situ observations. Figure 1   and speed, the geomagnetic aa, Kp x 10, and Dst indices, the Thule neutron monitor counting rate (pressure-corrected) and our assessment of the solar wind structure type, indicated at the beginning of the structure interval, where 2 is a high-speed stream, 3 is slow solar wind, ar.d 0 indicates that the structure type is unclear. Gray shading indicates a series of corotating high-speed streams. Vertical green lines indicate times of SSCs, typically associated during this interval with CIRs.
in 1967-1972 from IMPs 4 and 5) and neutron monitor data.  Figure 2 also illustrates how geomagnetic activity tends to be enhanced during the passage of high-speed streams, with the highest levels occurring in the vicinity of the CIRs, as is typical (e.g., Tsurutani et al. 2006). As noted above, such observations can help to indicate the presence of high-speed streams when limited solar wind observations are available. Similarly, the cosmic ray intensity may decrease during passage of a stream (Richardson 2004  3 ~ slow solar wind. During this more active interval, ICMEs and associated flows are more prominent than during the interval in Figure 2. The two vertical green lines indicate SSCs that are associated with 1he passage of shocks. Following both shocks, proton temperature (Tp) depressions relative to the "expected" value for Tp (T,,,,,, red graph) based on the observed solar wind speed are observed. Black shading indicates where Tp < O.ST exp' which is often indicative of an ICME. (See rc95 for further discussion of Texp. and its use in ICME identification.) The CMEassociated flows are also associated with enhanced geomagnetic activity, and cosmic ray Forbush decreases. Note that the ICME on 11-12 February 1969 is the prototypical "magnetic cloud", an ICME with an enhanced magnetic field that rotates smoothly in dL",ction suggestive ofa flux-rope like magnetic field, identified by Klein & Burlaga (1982).

Solar cycle variations in near-earth solar wind structures and parameters during 1963-2011
Having extended the solar wind classification from near present back to 1963, we now investigate how the solar wind structures and parameters near the Earth have varied over more than four solar cycles. Figure 4 shows the monthly sunspot number toge'.her with three-solar ro1ation averages during 1963-2011 of the percen1age of the time that the Earth was immersed in each type of solar wind flow, or whether the flow type was "unclear" (bottom panel). Note that we have been able to assess the structures that were present for much of cycle 20 using the combined da1a sets despite the incomplete OMNI2 dam coverage. (The larger occurrence of unclear periods from 1983 to 1994 reflects the intermittent solar wind coverage from IMP 8.) The results indicate that the percen1age of the time when the CME-associated flows are present tends to follow the solar activity/sunspot cycle, as would be expected given that the CME rate at the Sun follows the activity cycle (Webb & Howard 1994;Yashiro et aI. 2004;Robbrecht et aJ. 2009). CME-associated flows occupied up to -,!()...{j()% of the solar wind around solar maximum and were nearly absent (-5%) dming solar minimum. Interestingly, the occurrence of CMEassociated flows near the Earth does not appear to have been lower in the weaker sunspot cycles 20 and 23 compared to the larger cycle 21. (Cycle 22 is compromised by sigoificant "unclear" intervals although the occurrence of CME-associated flows is similar to the other cycles.) Coro1ating high-speed streams are most prevalent during the decline of the cycle where they may fonn --60% or more of the near-Earth solar wiud. They are however present at all s1ages of the cycle. Slow solar wind is also present throughout the solar cycle and appears to have been most prevalent (-70%) during the most recent solar minimum, associated with a corresponding fall in the presence of high-speed streams. (See Russell et aJ. 2010 for a discussion of some of the notable features of this solar minimum.) Figure 5 shows the variation in three-rotation averages of the aa index in all-solar-wiud and in CME-associated flows, coro1aring streams, and slow solar wind, in 1 %3-20 11. (Note the change of vertical scale in the bottom three panels.) We show aa (as in Richardson et aI. 2000Richardson et aI. , 2002 since it is the index with the longest time series (since 1868) and is often used fur long-tean studies of geomagnetic activity and its relationship to other phenomena (e.g., Love 2011 and references therein). Although aa tends to be enhanced at times of higher solar activity, note also the temponuy decreases around solar maximum indicated by arrows, such that some of the lowest levels of geomagnetic activity actually occur close to solar maximum. These features (and others to be discussed below) may be related to the lack of energetic solar phenomena near solar maximum, tenned the "Gnevyshev Gap" by Feminella & Storini 1997 who associate this with the temporary decrease in solar indices often found near sunspot maximum discussed by Gnevyshev (1967) and Gnevyshev (1977). (yVe note though that Kane 2005 has argued that the dip in aa in the Gnevyshev Gap does not strictly follow those in solar indices.) For a recent intelJlret.ation of the Gnevyshev Gap in terms of solar dynamo modelir.g, see Norton & Gallagher (2010). Another notable feature in Figure 5 is the unusually low values of aa (at least since 1963), in the recent extended minimum following cycle 23 (see also Tsurutani et aJ. 2011 ). These low values are evident in the averages for each solar wind flow type suggesting that they are pervasive throughout the solar wind. The all-solar-wind aa graph is overplotted in red on the stream-associated graph, illustrating how average values of aa tend to track those associated with streams, as previously noted by Richardson et al. (2002). However, average values fall below those in streams in the recent minimum because of the prominent contribution of weak activity in slow solar wind, as illustrated in Figure 6 which shows the contribution of each flow type to the three-rotation aa averages, including the unusually high HiO%) contribution from slow solar wind in 2009 right at solar minimum. Figure 6 also shows the CME-associated flo." contribution to aa that follows the solar activity cycle, and the stream-associated contribution that is most prominent during the declining phase/minimum. Another notable feature is that during the late declining phase of cycle 23, an increase in the '.::ontribution from CME-associated flows occurred in 2004-2006 together with a reduction in the stream-associated contribution that does not appear to have a counterpart in previous cycles. 100 .. 80 ..  oscillations that are well correlated with the direction of the IMF at Earth aIler allowing for the solar wind transit time to I AU (Scherrer et al. 1977). The root mean square of the daily measured fields is shown here. The second panel shows three-rotation averages of the aa geomagnetic index, while the bottom four panels show the interplanetary magnetic field intensity from 1964 to 2011, in all-solar-wind, CME-related solar wind, corotating high-speed streams, and slow solar wind; the all-solarwind average is overplotted in red in the lower panels. Cycles 21-23 show variations in the solar and interplanetary magnetic fields and the aa index that tend to follow the sunspot cycle. Clear structures that appear in each data set illustrate the close association between solar and interplanetary magnetic fields, and geomagnetic activity. Note in particular the ternponuy . A02-p5 decreases in solar and interplanrouy fields and geomagnetic activity around solar maximum in these cycles, indicated by arrows, in the Gnevyshev Gaps. As has been previously noted (e.g., Hedgecock 1975), cycle 20 did not show a clear increase in th, interplanetary magnetic field intensity (there are no Wil· cox solar magnetic field observations for comparison), and the magnetic fields are relatively weak in all·solar·wind regions. The generally lower values of aa during cycle 20 than in later cycles are also consistent with a weaker IMF.
The unusually low values of geomagnetic activity (fsurutani el al. 2011 ) and magnetic field strength (Smith & Balogh 2008;Connick et al. 2011 ) in the recent solar minimum are evident in Figure 7 and are observed in both slow solar wind and corotating streams. Field strengths within the few CME-associated flows observed in 200~20 I 0 are also weaker than those found during much of the period in Figure 7, indicating that the weaker fields durir.g this minimum were manifested in both transient and quasi'Slatioruuy solar wind flows. Both the recent minimum and cycle 20 confurm to the pattern previously discussed by Richardson el al. (2000Richardson el al. ( , 2002 in that the mean all·solar·wind [MF field strength closely tracks the mean fields foond in streams and slow solar wind. Our interpret.tion of this pattern, and of the remarkably similar variations in the solar and inter· planetmy magnetic fields, is that the variations in the average lMF intensity are closely related to solar magnetic field varia· tions, and are predominantly manifested in the background, non-transient solar wind. In particular, we emphasize that average fields at I AU, even during higher solar activity levels, are not dominated by the contribution of magnetic fields in transients that pass the observing spacecraft -the solar cycle vanation is essentially unchanged if the average field intensity is calculated using only the slow solar wind and stream intervals. . .

Percent Time CME·Assoc1ated Flows
[t has been suggesred, however (Owens & Crooker, 2006), that the solar cycle increases in the lMF strength arise fium closed field lines that are carried oUI by [CMEs to several AUs and are then opened by interchange reconnection. These field lines then arid to the open magnetic flux in the heliosphere, and contribute to (and cannot he distinguished fium) the background solar wind outside of the individual [CMEs thaI pass an observing spacecraft near the Earth. This model, using SOHOILASCO CME rates (http:/'cdaw.gsfc.nasa.gov/CME_list!) as input and a charactetistic time scale for reconnection of 50 days, can account for·the observed variation in the lMF during solar cycle 23 fitirly suc· cessfully. Unfortunately, there are no comparable CME observations to test the model for previous solar cycles, but we suggest thatthe occurrence ofCME-associated flows might provide a reasonable proxy fur the CME rate.
As noted above, the solar cycle variation in the IMF was much weaker in cycle 20 than in cycles 21-23. One possibility to account for this o~ation is that the CME rate was considerably lower in cycle 20 than in later cycles. However, the results in Figure .\ indicate that CMEs and associated flows were observed for similar frnctions of time near the Earth duro ing cycle 20 as in later cycles. This suggests that the CME rate was probably not substantially lower in cycle 20 but rather may have heen comparable to that in later cycles.
To examine this further. Figure 8 shows one-rotatiori averages of the interplanetary magnetic field intensity for all· solar-wind plotted ageins! the pen:entage of the time when CME-associated flows were present in each of the cycles 20-23. Assuming that the C~ociated flow occurrence is a reasonable proxy for the CME rate al the Sun (a caveat will be noted below), and the lMF strength is related to the magnetic flux added by [CMEs, we might then expect evidence of a LG. Richardson and H.V. ClPle: Near-earth solar wind flows and related geomagnetic activity positive correlation betwreD the IMF strength and the CM£. associated llow occurrence in each cycle, similar to that found between IMF intensity and LASCO CME rate in cycle 23 by Owens et aI. (2008)(cf. their Fig. I). The color of the line/symbol m each panel in Figure 8 indicates the time of observation.

While there is a genernl increase in IMF intensity for increasing
CME-associated flow occurrence in cycles 21-23 which may support the Owens & Crooker (2006) model, the distribution of points for cycle 20 is relatively flat because there was little increase in the IMF strength, despite 1he increase in the occur· rence of CME-associated flows, as solar activity levels increased. This observation would appear to pose a challenge to the Owens & Crooker (2006) model. One possibility is that the ICMEs in cycle 20 carried substantially less magnetic flux than in later cycles. The mean field in CME-associated flows (which include sheath regions as well as ICMEs) was indeed weaker (7.9 nT) during the maximum of cycle 20 compared with 9.1 nT in cycle 21, 10.\ nT in cycle 22, and 8.8 nT in cycle 23, but the difference seems too small to accOlmt for the near absence of a solar cycle field variation in cycle 20. The ICMEs might have had smaller volumes on average, and hence carried less magnetic flux, but presumably this would also have reduced the amount of time when CME-associated flows were presen~ which was not observed. A smaller rec?Mection time constant would contnbute to a smaller mag. nebc cycle (Owens & Crooker 2006), though it is not clear wby this should be a feature only of cycle 20.
A caveat to the results in Figure 8 is that it has been noted that the ICME rate at the Earth during cycle 23 did not track th~ CME rate at the Sun accurately (Riley et aI. 2006), so like-WIse, the occurrence of CME-associated flows during cycle 20 ' may also not fully reflect variations in the CME rate. On the other band, when the CME and ICME rates diverged in cycle 23, the CME rate actually rose more rapidly than the ICME rate. Hence, it is possible that the CME rate in cycle 20 similarly may have mcreased even more rapidly than is indicated by the CMEassociated flow occurrence at I AU, in which case the absence of the solar cycle variation in the IMF intensity is even more puzzling. Thus, in summary, we suggest that the observed increase in the occurrence of near-Earth CME-associated flows during cycle 20, indicative of an increase in the CME rate at the Sun, together with the weak increase in the interplanetary magnetic field strength during this cycle, may pose a challenge to Owens & Crooker (2006) proposal that solar cycle variations in the strength of the interplanetary magnetic field are associated with magnetic flux carried out by ICMEs. Figure 9 examines average solar wind speeds in 1963-20 II. We again show the monthly sunspot number and threerotation averaged aa index, together with three-rotation aver· ages of the solar wind speed for all-solar-wind and separately for corotating streams and CME-associated flows. The solar wind speed clearly sbows little correlation with solar activity levels. In f~ there is a tendency for local minima in the solar wind speed, including in CME-associated flows, in the Gnevyshev Gaps (mdicated by arrows) near solar maximum. Thus, some of the slowest solar winds during the solar cycle can occur close to solar maximum. The highest speeds teud to occur during the declining phase of the cycle when corotating streams are predominan~ but the persistence of these flows varies from cycle to cycle. In particular, in cycle 23, average flow speeds exceeding 500 kIn s -1 associated with oorotating streams were predominant only in 2003 but were present for -4 years in the dechne of cycle 20. Average solar wind speeds in the recent solar minimum were also evidently the lowest observed since  --~~ 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 Fie:. 9. Three-rotation averages of the solar wind speed in all-solarwind and in corotating streams and CME-associated flows during 1963-20 II are shown in the bottom three panels. 1be top panels show the monthly sunspot number and three-rotation averages of the aageomagnetic index. Arrows indicate local minima in the solar wind speed and geomagnetic activity near solar rnaximwn. the very. earliest in-situ obselVations. As also suggested by Tsurutaru et aI. (20 11), the above results indicate that the low geomagnetic activity levels in the recent minimum were a com· bination of low solar wind speeds (Fig. 9) due to the prevalence of slow solar wind at the expense of streams (Figs. 4 and 6), and weak interplanetary magnetic fields that are related to weak solar magnetic fields (Fig. 7). We also note that the CMErelated flows in cycle 23 were slower (.verages are below -500 Ian s-') during the ascending phase of the cycle than during the deseending phase, where average speeds were typically -5~ Ian s-'. This asymmetry may reflect the faster ambient solar wind speeds evident in Figure 9 and the increased treqoeney of fast (> \000 Ian s-') ICME-<!rlven intelJllanetary shocks noted by , during the declining phase of this cycle. Figure  and post-shock/sheath regions), corotating streams and slow solar wind, using the OMNI data and additional data sets, extending earlier studies of this type that commenced with observations in 1972 (Richardson et a!. 2000(Richardson et a!. , 2001(Richardson et a!. , 2002).
• The solar ("Sun as a star") and interplanetary magnetic field strengths in cycles 21-23 show variations that are similar in each data set, including temporary reductions close to solar maximum. Geomagnetic activity also shows similar reductions near solar maximum, and variations that are similar to those observed in the IMF and solar wind speed. • The low levels of geomagnetic activity during the recent solar minimum following cycle 23 are related to low solar wind speeds, due to a prevalence of slow solar wind mther than strearos and unusually weak interplanetary fields that are found in all·solar-wind flows and reflect 'weak solar magnetic fields. • The declining phase of cycle 23 is also characterized by an unusual persistence of CME-associated geomagnetic activity, extending to 2006, and a -I-year period in 2003 in which enhanced activity associated with streams was dominant, including some of the highest (three-rotation averaged) levels found during the study period. • Analysis of the weak solar cycle 20 suggests that CMEassociated flows were present for a similar fiaction of the time (-40%) as found in cycles 21-23, suggesting that the CME rate during this cycle was also comparable. The relatively weak increase in the interplanetary magnetic field intensity during cycle 20 may pose a problem for models of the solar cycle IMF variation that assume that field lines transported by ICMEs contribute to solar cycle variations in the IMF strength.