Properties and geoeffectiveness of solar wind high-speed streams and stream interaction regions during solar cycles 23 and 24
CConfidential manuscript submitted to
JGR-Space Physics
Properties and geoeffectiveness of solar wind high-speed streamsand stream interaction regions during solar cycles 23 and 24
Maxime Grandin , , Anita T. Aikio , and Alexander Kozlovsky Sodankylä Geophysical Observatory, University of Oulu, Sodankylä, Finland Department of Physics, University of Helsinki, Helsinki, Finland Ionospheric Physics Unit, University of Oulu, Oulu, Finland
Key Points: • Stream interaction regions and high-speed streams (SIR/HSSs) are 20–40% less geo-effective during solar cycle (SC) 24 than during SC23 • The most geoeffective SIR/HSSs in solar cycles 23 and 24 take place in the earlydeclining phases • During the late declining phase of SC23, both SIR/HSS event number and maximumvelocity are highest, yet their geoeffectiveness is low
Corresponding author: Maxime Grandin, [email protected] –1– a r X i v : . [ phy s i c s . s p ace - ph ] J un onfidential manuscript submitted to JGR-Space Physics
Abstract
We study the properties and geoeffectiveness of solar wind high-speed streams (HSSs) ema-nating from coronal holes and associated with stream interaction regions (SIRs). This paperpresents a statistical study of 588 SIR/HSS events with solar wind speed at 1 AU exceeding500 km/s during 1995–2017, encompassing the decline of solar cycle 22 to the decline ofcycle 24. Events are detected using measurements of the solar wind speed and the inter-planetary magnetic field (IMF). Events misidentified as or interacting with interplanetarycoronal mass ejections (ICMEs) are removed by comparison with an existing ICME list.Using this SIR/HSS event catalog (list given in the supplementary material), a superposedepoch analysis of key solar wind parameters is carried out. It is found that the number ofSIR/HSSs peaks during the late declining phase of solar cycle (SC) 23, as does their velocity,but that their geoeffectiveness in terms of the AE and SY M – H indices is low. This can beexplained by the anomalously low values of magnetic field during the extended solar mini-mum. Within SC23 and SC24, the highest geoeffectiveness of SIR/HSSs takes place duringthe early declining phases. Geoeffectiveness of SIR/HSSs continues to be up to 40% lowerduring SC24 than SC23, which can be explained by the solar wind properties. Corotating solar wind high-speed streams (HSSs) were discovered by
Snyder et al. [1963], who showed a correlation between recurring geomagnetic activity and the solarwind speed measured by Mariner 2. High-speed streams were later associated with solarcoronal holes [e.g.,
Krieger et al. , 1973;
Sheeley Jr et al. , 1976]. Since a given coronal holemay persist for several solar rotations, this will lead to recurring geomagnetic activity with a ∼ Chkhetiia , 1975;
Hapgood , 1993;
Temmer et al. , 2007;
Crowley et al. , 2008], this periodicitycorresponding to the synodic solar rotation period as viewed from Earth. A HSS is charac-terized by an enhancement in the solar wind velocity lasting for several days. Typically, thespeed exceeds 500 km/s for two to three days, and may reach a maximum above 800 km/s[e.g.,
Denton and Borovsky , 2012;
Kavanagh et al. , 2012].When the high-speed stream overtakes the slow-speed background stream, this leadsto compressions of both the interplanetary magnetic field (IMF) and the plasma density inthe rising-speed portion of the high-speed stream [
Belcher and Davis , 1971]. The resultingstructure was called corotating interaction region (CIR) [
Smith and Wolfe , 1976;
Gosling andPizzo , 1999]. In this paper, we refer to this interaction region as “stream interaction region(SIR)” regardless of the number of solar rotations during which it is observed. This practiceis similar to, e.g.,
Richardson [2018]. In addition, the high-speed stream events that we studycontain the interaction regions in the leading edge of the streams, so we refer to these eventsas SIR/HSS.SIR/HSSs have been particularly studied during the declining phase of the 11-yearsolar cycle, since at that time they represent the main cause for geomagnetic disturbances[
Gonzalez et al. , 1999;
Tsurutani et al. , 2006]. On the other hand, during the maximum yearsof solar activity, coronal mass ejections (CMEs) are the major cause of geomagnetic distur-bances [e.g.,
Richardson and Cane , 2012a]. CMEs are eruptions of magnetized plasma fromthe solar atmosphere, with a broad range of propagation speeds [
Gopalswamy and Kundu ,1992]. Interplanetary coronal mass ejections (ICMEs) are the interplanetary counterpart ofCMEs [e.g.,
Schwenn , 1983;
Sheeley et al. , 1985;
Lindsay et al. , 1999;
Webb et al. , 2000].Those produce the most extreme space weather events [e.g.,
Richardson et al. , 2000, 2001;
Zhang et al. , 2007;
Echer et al. , 2013].It is well-established that SIR/HSSs are predominantly responsible for weak to mod-erate geomagnetic storms [e.g.,
Tsurutani et al. , 2006].
Zhang et al. [2008] found that about80% of the geomagnetic storms produced by “pure” SIR/HSSs (i.e., SIR/HSSs which arenot interacting with an ICME) are weak ( − < Dst ≤ −
30 nT) to moderate ( − < –2–onfidential manuscript submitted to JGR-Space Physics
Dst ≤ −
50 nT), while
Alves et al. [2006] showed that only about 33% of the SIR/HSSs areresponsible for moderate to intense storms (Dst ≤ −
100 nT).
Chi et al. [2018] found thata large fraction of the most intense SIR/HSS-related geomagnetic storms are produced bySIR/HSSs interacting with an ICME. According to
Kilpua et al. [2017], although SIR/HSSsare frequent during solar minimum, they produce fewer geomagnetic storms as they aregenerally characterized with lower speed gradients, lower dynamical pressure peaks andlower interplanetary magnetic field magnitude. Yet,
Richardson and Cane [2012a] revealedthat almost half of the large storms during solar minimum are due to SIR/HSSs.Statistical studies of SIR/HSS effects on the magnetosphere–ionosphere–thermospheresystem have shown dropouts of relativistic electrons in the outer radiation belt [
Borovsky andDenton , 2009;
Morley et al. , 2010], followed by a recovery to higher flux levels in the case ofstrong events [
Denton and Borovsky , 2012], enhancement of substorm activity [
Tsurutaniet al. , 2006], depletion of the ionospheric F region [ Denton et al. , 2009;
Grandin et al. ,2015], energetic ( E >
30 keV) electron precipitation into the D region [ Meredith et al. , 2011;
Grandin et al. , 2017], and enhancement of ULF wave activity in the magnetosphere [
Mathieand Mann , 2001].Because of the strong influence of SIR/HSS events on the planetary environment,objective criteria to identify SIR/HSS events from solar wind data would be very useful.Previous statistical studies have used methods to detect SIR/HSSs focusing on stream in-terfaces, and hence searching for not only enhancement in the solar wind radial velocitybut also a west–east deflection of the plasma flow [e.g.,
Morley et al. , 2010;
Denton andBorovsky , 2012;
Kavanagh et al. , 2012]. Those methods often required a visual inspectionof the events and/or comparison with existing lists of other solar wind disturbances suchas ICMEs to remove false positives. For instance,
Jian et al. [2006] produced a fairly com-prehensive list of events during 1995–2004, which was extended till the end of 2016 by
Chi et al. [2018]. The events were picked by eye based on criteria on the solar wind speedand total perpendicular pressure. The starting time corresponds to the time when the totalperpendicular pressure starts increasing, which is close to the time when the speed also startsincreasing.
Richardson and Cane [2012b] have also produced a classification of solar windflows, including an identification of SIR/HSSs, during 1963–2011 based on visual inspectionof solar wind data as well as geomagnetic observations, energetic particle data, and neutronmonitor data. Others have developed automated methods using only criteria related to theenhancement of the solar wind speed, such as Maris , Muntean et al. ( ) during 2009–2016 or
Gupta and Badruddin [2010] during 1996–2007.More generally, attempts to automatically identify solar wind flow types between inter-stream flow (slow wind), coronal-hole flow (high-speed streams), and transient (ICMEs) havebeen presented by, e.g.,
Zhao et al. [2009] during 1998–2008 from ACE data,
Reisenfeldet al. [2013] during the Genesis mission (2001–2004), and
Xu and Borovsky [2015] during1963–2013 from OMNI data. However, when comparing results from the latter three identifi-cation schemes,
Neugebauer et al. [2016] found that the overall agreement is limited, with allthree algorithms giving a same flow type only 49% of the time.This paper presents a method to identify SIR-associated HSSs, which is applied to 23years of solar wind observations from 1995 until 2017. This detection method is based onthe one used to obtain the HSS events analyzed in
Grandin et al. [2015, 2017] for 2006–2008, and for this paper some improvements have been developed, detailed in section 3.The aim of this method is to find the time when the SIR region with compressed magneticfield hits the bow shock. Solar wind speed reaches the maximum typically 1–2 days afterthis zero epoch. It was shown in
Grandin et al. [2015, 2017] that the zero epoch selectedin this manner corresponds very well to the time when geomagnetic activity in terms of theAE index starts to increase. The list of SIR/HSSs obtained with this new algorithm is usedto self-consistently study the features of SIR/HSSs impinging on the Earth’s magnetopauseduring the different phases of two solar cycles, and in specific their geoeffectiveness. –3–onfidential manuscript submitted to
JGR-Space Physics
This paper is organized as follows: the data sets used in this study are presented insection 2, and the algorithm developed to gather the list of SIR/HSS events is described insection 3. Section 4 presents the results on SIR/HSS characteristics from the end of solarcycle 22 (SC22) until late 2017 (declining phase of SC24). Section 5 discusses the resultsand compares the obtained SIR/HSS list with other existing ones, and section 6 summarizesthe main conclusions.
The data used in this study consist of interplanetary magnetic field (IMF) and solarwind data measured near the Earth between 1995 and 2017. The data were obtained by sev-eral satellites of which the main ones are WIND (1995–present) [
Ogilvie et al. , 1995] andACE (1998–present) [
McComas et al. , 1998;
Smith et al. , 1998]. The measured data areavailable through the OMNI database [
King and Papitashvili , 2005], where they have alreadybeen propagated to the terrestrial bow shock. The parameters of interest in this study arethe IMF magnitude B , the solar wind speed V , the solar wind density N , and the Akasofu ε parameter [ Akasofu , 1979] calculated using ε = πµ V B L sin (cid:18) θ (cid:19) , (1)with µ the vacuum permeability, θ the IMF clock angle in the plane perpendicular to theSun–Earth direction, and L (cid:39) R E . The Akasofu ε parameter is a coupling function oftenused as a proxy for energy input from the solar wind into the magnetosphere. For this study,the OMNI data were retrieved for years 1995–2017 at 1 h time resolution.We also take the 1–h averages of the auroral electrojet ( AE ) and SYM–H geomagneticindices [
Davis and Sugiura , 1966;
Iyemori , 1990] from OMNI to monitor the response of thegeospace in terms of substorm and storm activity, respectively.In addition, in order to discuss the properties of SIR/HSSs during different phases ofsolar cycles 23 and 24, the monthly values of the revised sunspot number between 1995 and2017 have been gathered from the World Data Center Sunspot Index and Long-Term SolarObservations (SILSO) [
SILSO World Data Center , 1995-2017].
Our aim is to find SIR/HSS events that can be geoeffective. Therefore, for maximumvelocity we have chosen a threshold of 500 km/s. One can see that the Akasofu epsilon so-lar wind coupling function (eq. 1) depends linearly on solar wind velocity. The papers by
Kavanagh et al. [2012] and
Denton and Borovsky [2012] are also in accordance with thisselection, see point 2(c) below. However, the algorithm presented below could be adapted toalso detect SIR/HSSs of lower maximum speed.The algorithm used to detect SIR/HSS events for this study consists of four steps. Thefirst three steps each consider a criterion based on solar wind data at 1 h resolution providedby OMNI.1. The SIR is formed by the plasma compression arising from the interaction of theslow and fast solar wind streams. Therefore the leading edge of SIRs exhibits anenhancement in the IMF magnitude B [e.g., Richter and Luttrell , 1986;
McPherronand Weygand , 2006;
Denton and Borovsky , 2012]. Hence, as a first criterion, the timederivative of B is estimated at each time t i usingd B d t (cid:39) ∆ t ( B i + − B i − ) , (2)with ∆ t = t i + − t i = t i at which this time derivative value exceeds the empirically determined threshold –4–onfidential manuscript submitted to JGR-Space Physics of 0.6 nT/h is then flagged as a candidate SIR/HSS event starting time ( t ). If thethreshold value is too large, only SIRs that are associated with a forward shock willbe detected. Jian et al. [2006] found that, on average, the occurrence rate of forwardshocks in SIRs at 1 AU is about 18%. However, if the threshold value is too small,then we are no more looking at compressed plasma, but some natural fluctuations inB. The threshold value used in this paper was found empirically, and comparison toother existing SIR lists described in Section 5.1 shows that the threshold works well,since no events are missed due to this criterion.2. The second criterion is based on the solar wind velocity V . The candidate eventsobtained with criterion (1) are tested for three characteristics:(a) To avoid detecting events taking place while the background solar wind speed isalready high, at time t , one must have V ( t ) ≤
450 km/s, bearing in mind thatoccasional compound events consisting of two high-speed streams interacting witheach other might not be detected due to this restriction. This value of 450 km/swas also determined empirically by visual inspection of the detected/undetectedSIR/HSS events when trying various velocity thresholds.(b) To ensure that the velocity starts increasing shortly after the beginning of the event,the average velocity slope between t and t + Bame et al. [1976] and
Gosling et al. [1976].(c) To keep only events with solar wind speed significantly higher than the slow wind,a threshold is defined: the solar wind speed must reach at least 500 km/s within[ t , t + Kavanagh et al. [2012] in their superposed epoch study of >
30 keV precipitation during high-speedstreams, and falls between the thresholds defined by
Denton and Borovsky [2012]for “weak” (450 km/s) and “strong” (550 km/s) HSS events. It is also consistentwith the retained velocity thresholds used for the detection of coronal-hole flow bythe Genesis spacecraft mission [
Reisenfeld et al. , 2013].The candidate SIR/HSS event starting times which verify each of those three condi-tions are kept for the next step.3. The third criterion makes sure that one does not detect a same SIR/HSS event twicewhen, for instance, criteria (1) and (2) are met at several times during the event. Crite-rion (3) consists in keeping only SIR/HSS event starting times separated by at least3 days. Beginning with the first detected SIR/HSS event at t , all candidate timesbetween t and t + t (cid:48) , is retained while all others are removed from the list. Thisguarantees that the solar wind speed starts increasing immediately after the retainedstarting time. This operation is repeated for the next SIR/HSS candidate occurringafter t (cid:48) + Richardson and Cane [2010], which covers years from 1996 onward and is regularlyupdated ( ). A candidate is considered a likely ICME or contaminated by an ICME if oneevent from the Richardson and Cane list with V max ≥
500 km/s reached the Earthbetween t − t + t being the date of the candidate SIR/HSS event.In total, 122 events matching criterion 4 were found and were removed from the SIR/HSScatalogue. As a consequence, this leaves a total of 588 SIR/HSS events obtained with thisfour-step method. This list of SIR/HSS events is provided as supporting information. The list –5–onfidential manuscript submitted to JGR-Space Physics gives for each SIR/HSS event its starting time t , initial solar wind speed, the time and valueof maximum solar wind speed during [ t , t +3 days], and an end time defined as the firstoccurrence when the solar wind speed drops below 450 km/s after reaching its maximum.All times are given in UT for solar wind propagated at the terrestrial bow shock as given inOMNI. It should be noted that this list does not claim to be comprehensive, as (i) this studyfocuses on SIR/HSS events whose velocity reaches at least 500 km/s, (ii) occasionally, someSIR/HSS events with V max ≥
500 km/s might have been undetected by the method, and (iii)events containing ICMEs during [ t − t + x -axis. Figure 1ashows the solar wind speed. It is clear that the beginning of each event is set very close to thetime when the solar wind velocity starts to increase, which naturally comes from criteria 2band 3. The horizontal red dashed line indicates the threshold value at 500 km/s for a HSSevent (criterion 2c). Figure 1b gives the IMF magnitude | B | which is used for criterion 1.The sharp time derivative used for the detection is particularly prominent at the beginningof the second event. From Figure 1c, which shows the solar wind number density, it can benoted that the beginning of the detected SIR/HSS events is close to the maximum of the den-sity peak (again, this is more prominent for the second event), which corresponds to earlierfindings [e.g., Gosling et al. , 1972;
McPherron and Weygand , 2006;
Denton et al. , 2009].Figure 1d displays the AE index, which is a proxy for substorm activity [ Iyemori , 1990].Substorm activity starts to increase very close to the determined onset time of the events ineach case. It is interesting to note that AE starts to increase well before the solar wind veloc-ity has reached a high value, which indicates that the SIR is geoeffective. Figure 1e displaysthe
SYM–H index, which describes the intensity of the storm time ring current [
Wanliss andShowalter , 2006]. In the 14 June event, which is associated with the strongest storm, the zeroepoch matches very well with the sudden storm commencement signature showing as a pos-itive peak in
SYM–H , likely to be produced by the slight dayside magnetopause compressionassociated with the SIR arrival. A smaller positive peak is observed also for the other events,except the first one on 6 June. All these events are associated with weak magnetic storms. Itis well-known that SIR-associated geomagnetic storms are frequent in the declining phase ofthe solar cycle and are typically weak to moderate [
Tsurutani et al. , 2006;
McPherron andWeygand , 2006;
Richardson and Cane , 2012a;
Grandin et al. , 2017].Figure 2 shows the same data as Figure 1 but for the whole year 2008, during which33 SIR/HSS events were detected by the algorithm. From this longer time series, the featuresdescribed above can be identified in most of the detected events, namely the fact that thebeginning of events corresponds well with the start of the solar wind speed increase, witha peak in the solar wind density within the SIR region, with the start of AE index enhance-ment, and is near the sudden storm commencement signature when it is present. One cannote that the event contaminated by the only ICME with V max ≥
500 km/s in 2008, whichtook place on 4 December (magenta triangle in Figure 2a), was rejected by the algorithmbecause of criterion 4.Figure 3 gives the yearly distribution of ICMEs with V max ≥
500 km/s from the Richard-son and Cane list between 1996 and 2017 (the envelope of the histogram). Orange bars showthe ICME events that were identified as SIR/HSSs by our algorithm with steps 1–3, beforethey were removed based on criterion 4, and blue bars show the remaining ICME events. Theblack line gives the monthly sunspot number during those years. It is noteworthy that notall the ICMEs are misidentified as SIR/HSSs by our algorithm, only 122 out of 237 ICMEswith speeds larger than 500 km/s, corresponding to about 50%. Of all the detected SIR/HSSevents, the amount of rejected events (ICMEs or SIR/HSSs contamined by ICMEs) is 17%(122/709). This underlines the fact that comparison with an ICME list (criterion 4) is a nec-essary step in the algorithm, as the solar wind parameters used in the identification scheme –6–onfidential manuscript submitted to
JGR-Space Physics
Figure 1.
SIR/HSS events detected by the algorithm in June 2008 with a maximum solar wind velocitygreater than 500 km/s. Each vertical dashed line corresponds to the beginning of an event. The time seriesshown are (a) the solar wind velocity, (b) the IMF magnitude, (c) the solar wind number density, (d) the AE index, and (e) the SYM–H index. The horizontal dashed red line in the first panel indicates V =
500 km/s,used for criterion 2(c). V ( k m / s ) (a) | B | ( n T ) (b) N ( c m - ) (c) AE ( n T ) (d) / /
01 24 / / /
02 27 /
02 08 /
03 22 / /
03 03 /
04 16 / / /
04 19 /
05 28 /
05 06 /
06 14 / / /
06 10 /
07 20 /
07 09 / /
08 03 /
09 14 /
09 30 /
09 10 /
10 26 /
10 06 /
11 15 /
11 24 /
11 22 /
12 30 / HSS starting date -50-2502550 SY M - H ( n T ) (e) Figure 2.
Same as Figure 1 but showing the data and detected events for the entire year 2008. The magentatriangle in panel (a) corresponds to the ICME falsely detected as SIR/HSS in 2008, which was removed whenapplying criterion 4 (see Figure 3). –7–onfidential manuscript submitted to
JGR-Space Physics N u m be r o f e v en t s Year S un s po t nu m be r SC23 SC24
Figure 3.
Histogram, left axis: Yearly distribution of ICMEs from
Richardson and Cane [2010] (updatedonline version of January 2019) with V max ≥
500 km/s during 1996–2017 (envelope) and, among these,ICMEs misidentified as SIR/HSSs by the algorithm or ICMEs embedded in a SIR/HSS (orange). Black line,right axis: Monthly sunspot number during those years. during ICMEs quite often exhibit similar features as SIR/HSSs. This issue is not severeduring solar minimum years, but becomes important during solar maximum years. Onemust also keep in mind that some of the rejected events may be real SIR/HSS events thatare interacting with a simultaneously occurring ICME event near the ecliptic plane [see, e.g.,
Al-Shakarchi and Morgan , 2018;
Shugay et al. , 2018].
In order to provide an overview of all the detected events during 1995–2017, Fig-ure 4 displays the solar wind velocity plotted as a function of year ( y -axis) and day of so-lar rotation ( x -axis), when using Bartels rotations (27-day period). The time series startsfrom 1 January 1995 on day 17 of Bartels rotation 2204 and follows the horizontal axis un-til day 27, and then continues above from day 0 of Bartels rotation 2205 which starts on11 January, and so on. The y -axis however indicates year instead of Bartels rotation num-ber for an easier comparison with the other figures. This way of displaying the solar windspeed enables one to identify features recurring over several solar rotations, which appearas vertical structures in the plot. Two horizontal magenta dashed lines, in August 1996 andDecember 2008, indicate the transitions between solar cycles 22 and 23, and solar cycles 23and 24, respectively, obtained from ftp://ftp.ngdc.noaa.gov/STP/space-weather/solar-data/solar-indices/sunspot-numbers/cycle-data/table_cycle-dates_maximum-minimum.txt . The starting dates of the detected SIR/HSS events are indicated aswhite filled circles on top of the solar wind velocity color plot. Several features indicatingrecurring elevated solar wind velocities can be identified, in particular during year 1995, –8–onfidential manuscript submitted to JGR-Space Physics
Figure 4.
Color plot: Solar wind velocity plotted as a function of day number within 27-day Bartelsrotation ( x -axis, 1 Jan 1995 corresponds to day 17 of Bartels rotation 2204) and year from 1995 to2017 ( y -axis). White circles: starting dates of the detected SIR/HSS leading edges. The magentadashed lines indicate the boundaries between solar cycles obtained from ftp://ftp.ngdc.noaa.gov/STP/space-weather/solar-data/solar-indices/sunspot-numbers/cycle-data/table_cycle-dates_maximum-minimum.txt . –9–onfidential manuscript submitted to JGR-Space Physics years 2003–2004, years 2003–2008, and from 2015 onwards. During one solar rotationperiod, there may be several recurring intervals of high-speed flows. Discussion of the originof these long-lived recurring high-speed streams is beyond the scope of the present paper,but one can note that, e.g.,
Temmer et al. [2007] suggest that the 9-day period in solar windparameters, showing up as higher harmonic of the solar rotation frequency, is caused bythe “periodic” longitudinal distribution of coronal holes on the Sun recurring for severalsolar rotations, especially during January–September 2005. This 9-day periodicity can beseen in Figure 4 during that time period, with enhanced solar wind speed recurring aroundBartels rotation day 0, day 9, and day 18. It is noteworthy that structures from which high-speed solar wind flows may remain present at a same phase of the Bartels rotation for upto several years (see, e.g., the high-velocity signatures near the longitude corresponding today 10, recurring from early 2005 until late 2006, or those near the longitude correspond-ing to day 2, recurring from mid 2006 until late 2008). This is consistent with findings by
Heidrich-Meisner et al. [2017] who studied a same coronal hole structure for twelve solar ro-tations in 2006. The resulting HSS reached the Earth eleven times, yet exhibiting variabilityin its signature observed by ACE at L1 as the spacecraft mapping drifted to different regionswithin the coronal hole. On the other hand,
Krista et al. [2018] tracked equatorial coronalholes at the surface of the Sun during 2011–2014 using the unique 360 ◦ coverage offered bythe configuration of the Solar–Terrestrial Relations Observatory (STEREO) constellation andthe Solar Dynamics Observatory (SDO). During those years in the maximum phase of SC24,most equatorial coronal holes had a lifetime of less than 100 days ( < Borovsky and Denton [2006] for the endof SC22 and the beginning of SC23. The fading of the recurring coronal hole signaturescoincides quite strikingly with the transition between solar cycles 23 and 24 in late 2008,after which HSSs become less frequent for a couple of years.Figure 5 shows the yearly number of SIR/HSS events with V max ≥
500 km/s detectedby our method as a histogram plot, where all the events that are contaminated by ICMEshave been removed. The sunspot number is shown as a colored curve, which corresponds tofour phases of solar cycles which we will consider below (in particular in section 4.2): risingphase (red), maximum (green), early declining phase (blue), and late declining phase (black).The selected time limits for the phases are given in Table 1. The criterion used to separatethe rising, maximum and declining phases was that, for both cycles 23 and 24, the doublepeak of the monthly sunspot number be included in the maximum phase. The motivation forsubdividing the declining phase of solar cycles into early and late parts was to obtain solarcycle divisions of roughly similar durations, and to reveal potential differences within thedeclining phase itself.While it is well-known that high-speed streams are frequent during the declining phasesof solar cycles [e.g.,
Gonzalez et al. , 1999], many events are also detected during the otherphases of solar cycles 23 and 24, which is consistent with reports by
Richardson and Cane [2012b] during solar cycles 20 to 23. It is noteworthy that even the rising phases contain alarge number of SIR/HSS events, corresponding to 69% and 48% of yearly numbers dur-ing the declining phases of SC23 (rising-phase years: 1997–1999; declining-phase years:2003–2008) and SC24 (rising-phase years: 2009–2010; declining-phase years: 2015–2017),respectively. In fact, as was noted by
Echer et al. [2013], SIR/HSS events were responsiblefor 30% of the moderate geomagnetic storms taking place during the rising phase of SC23.The following common characteristics appear to repeat during the two observed cycles.After the sunspot minimum is reached, the number of SIR/HSS events is at minimum in thebeginning of a new cycle both during cycles 23 (year 1997) and 24 (year 2009), which is –10–onfidential manuscript submitted to
JGR-Space Physics N u m be r o f e v en t s Year S un s po t nu m be r SC23 SC24
Figure 5.
Histogram: Yearly distribution of high-speed stream events detected with the method using thecriterion V max ≥
500 km/s. Line plot: Monthly sunspot number. The colors of the curve indicate the timeintervals which are referred to as “rising phase” (red), “maximum” (green), “early declining phase” (blue) and“late declining phase” (black) of solar cycles.
Table 1.
Subdivision of solar cycles (SC) between 1995 and 2017
Phase Start date End date Duration (months)
SC 22 – late declining Jan 1995 Jul 1996 19SC 23 – rising Aug 1996 Dec 1999 41SC 23 – maximum Jan 2000 Jun 2002 30SC 23 – early declining Jul 2002 Dec 2005 42SC 23 – late declining Jan 2006 Nov 2008 35SC 24 – rising Dec 2008 Aug 2011 33SC 24 – maximum Sep 2011 Jun 2014 34SC 24 – early declining Jul 2014 Dec 2017 42 –11–onfidential manuscript submitted to
JGR-Space Physics
Yearly distribution of detected HSS events N u m be r o f e v en t s Year S un s po t nu m be r Figure 6.
Histogram: Yearly distribution of high-speed stream events detected with the method and binnedaccording to their maximum solar wind speed value. Line plot: Monthly sunspot number. consistent with findings reported by
Jian et al. [2011]. Then the number of SIR/HSSs startsto increase during the next two years in the rising phase. During sunspot maximum yearsthe increase stops and there is a local minimum (year 2000 for cycle 23 and year 2014 forcycle 24). Even if one added the HSSs rejected because they contain an embedded ICME(part of the orange bars in Figure 3), the number of SIR/HSS events would still drop and alocal minimum would still be found during these years (2000 and 2014). This confirms theobservation made by
Jian et al. [2006] during the maximum phase of solar cycle 23.In the early declining phase of solar cycles, the number of SIR/HSS events starts toincrease again and the maximum number of events occurs obviously during the last year ofcycle 22 (1996, but we have data only for the last two years) and the second last year (2007)for cycle 23. For cycle 24, the number of events has almost the same values for the last twoobserved years, i.e., 2016 and 2017. If the length of cycle 24 were the nominal 11 years, thiswould imply that sunspot minimum would take place in 2019 and we could expect to havethe maximum number of SIR/HSS events in year 2018 or 2019. The number of SIR/HSSevents in 2007 is 39, and there were 35 events in years 1996 and 2016. These numbers cor-respond roughly to 3 SIR/HSS events per month with maximum speeds exceeding 500 km/sduring maximum occurrence.One interesting characteristic of the detected SIR/HSS events is the distribution oftheir maximum solar wind speed. Figure 6 gives the yearly distribution of the 588 detectedSIR/HSS events, binned according to their maximum solar wind speed value during the firstthree days in the 1 h resolution data. The following clear features can be observed. First,during the first full year of a new cycle, when the number of SIR/HSS events is very low,the velocities of the HSSs are also low (years 1997 and 2009) with maximum velocitiesremaining mainly below 600 km/s. Second, during the declining phases, when the numberof SIR/HSS event maximizes, the proportion of events with velocities higher than 650 km/salso increases. Third, HSS velocities larger than 750 km/s are very rare and occur typically –12–onfidential manuscript submitted to
JGR-Space Physics -2 0 2 4
Time from zero epoch (days) V ( k m / s ) SC22 - late declining (a) N=45 -2 0 2 4
Time from zero epoch (days)
SC23 - rising (b) N=84 -2 0 2 4
Time from zero epoch (days)
SC23 - maximum (c) N=44 -2 0 2 4
Time from zero epoch (days)
SC23 - early declining (d) N=100 -2 0 2 4
Time from zero epoch (days) V ( k m / s ) SC23 - late declining (e) N=106 -2 0 2 4
Time from zero epoch (days)
SC24 - rising (f) N=56 -2 0 2 4
Time from zero epoch (days)
SC24 - maximum (g) N=46 -2 0 2 4
Time from zero epoch (days)
SC24 - early declining (h) N=106
Figure 7.
Superposed epoch analysis of the solar wind speed during SIR/HSS events during (a) the latedeclining phase of SC22, (b) the rising phase of SC23, (c) the maximum of SC 23, (d) the early decliningphase of SC23, (e) the late declining phase of SC23, (f) the rising phase of SC24, (g) the maximum of SC24,and (h) the early declining phase of SC24. The number indicated in the bottom-right-hand corner of eachpanel corresponds to the number of SIR/HSS events during the corresponding solar cycle subdivision (cf.Figure 5).
A comparison of SIR/HSS features during the solar cycle phases defined in Figure 5 ismade using the superposed epoch analysis method, which was used in
Grandin et al. [2015,2017]. The superposed epoch method is a statistical analysis invented by
Chree [1913],based on the collection of a large number of events, the definition of a reference time called“zero epoch” in the event timeline and the extraction of statistical properties (e.g., mean,median, and standard deviation) of the physical parameters of the events as a function of timerelative to the zero epoch [see also 3.1 in
Grandin , 2017]. Figures 7 to 12 show the statisticalbehaviors of key solar wind parameters and geomagnetic indices during high-speed streamevents taking place at each solar cycle phase.Figure 7 shows the superposed epoch analysis of the solar wind velocity during theSIR/HSS events detected during (a) the late declining phase of SC22, the (b) rising, (c) max-imum, (d) early declining and (e) late declining phases of SC23, and the (f) rising, (g) max-imum and (h) early declining phases of SC24. In each panel, median values are shown inred, and upper and lower quartiles in blue. The data are plotted from two days before thezero epoch until five days after, the zero epoch of each event corresponding to its startingtime given by the algorithm. The number of events used to obtain the curves in each panel isgiven in the bottom-right-hand corner of the panel. –13–onfidential manuscript submitted to
JGR-Space Physics -2 0 2 4
Time from zero epoch (days) B ( n T ) SC22 - late declining (a) N=45 -2 0 2 4
Time from zero epoch (days)
SC23 - rising (b) N=84 -2 0 2 4
Time from zero epoch (days)
SC23 - maximum (c) N=44 -2 0 2 4
Time from zero epoch (days)
SC23 - early declining (d) N=100 -2 0 2 4
Time from zero epoch (days) B ( n T ) SC23 - late declining (e) N=106 -2 0 2 4
Time from zero epoch (days)
SC24 - rising (f) N=56 -2 0 2 4
Time from zero epoch (days)
SC24 - maximum (g) N=46 -2 0 2 4
Time from zero epoch (days)
SC24 - early declining (h) N=106
Figure 8.
Same as Figure 7 for the IMF magnitude.
Out of these eight panels, the highest median velocities take place during the late de-clining phase of SC23 covering years 2006–2008 (panel e). The maximum median velocityin these superposed epoch plots reaches values of ≥
550 m/s from the end of day 2 to thebeginning of day 3 (please note that values 0–1 in the x axis correspond to day 1 and valuesfrom 1 to 2 to day 2). Second largest velocities are found in the early declining phases ofSC23 and SC24 (panels d and h). Clearly smallest velocities occur in the rising phases ofsolar cycles (panels b and f).It looks a bit surprising that the velocity in the late declining phase of SC22 is notshowing particularly high values, unlike in SC23. The answer may be related to the specialcharacteristics of the long-lasting solar minimum at the end of SC23. The study of coronalholes by de Toma [2011] provides insight in this topic. According to de Toma [2011], duringthe 1996 solar minimum, the two large polar coronal holes were the dominant coronal holeson the Sun and stable sources of fast solar wind over the poles, like in typical solar minimumconditions. This gave a relatively slow solar wind at the Earth, originating mostly from theedges of the polar coronal holes. However, the long period of low solar activity from 2006to 2009 was characterized by weak polar magnetic fields and polar coronal holes smallerthan observed during the previous minimum. Instead, large low-latitude coronal holes werepresent on the Sun until 2008 and remained important sources of recurrent high-speed solarwind streams. By early 2009, most the low-latitude coronal holes had closed down. Withthe increase of the new SC24 activity, small, mid-latitude coronal holes appeared. The fastwind in the rising phase was coming mostly from the edges of the polar coronal holes andoccasionally from the small, mid-latitude coronal holes. Finally, the high-speed wind in thedeclining phases comes typically from the low-latitude coronal holes [ de Toma , 2011].Figure 8 shows the IMF magnitude B in the same format. The largest values for B inthe SIR are found in the early declining and maximum phases of SC23 with the median val-ues exceeding 10 nT during day 1. Inside SC24, the largest values of B take place in the earlydeclining phase with median values reaching 9 nT. This is a confirmation that the maximum B value in SIRs does not simply follow solar activity given by the monthly sunspot number, –14–onfidential manuscript submitted to JGR-Space Physics -2 0 2 4
Time from zero epoch (days) N ( c m - ) SC22 - late declining (a) N=45 -2 0 2 4
Time from zero epoch (days)
SC23 - rising (b) N=84 -2 0 2 4
Time from zero epoch (days)
SC23 - maximum (c) N=44 -2 0 2 4
Time from zero epoch (days)
SC23 - early declining (d) N=100 -2 0 2 4
Time from zero epoch (days) N ( c m - ) SC23 - late declining (e) N=106 -2 0 2 4
Time from zero epoch (days)
SC24 - rising (f) N=56 -2 0 2 4
Time from zero epoch (days)
SC24 - maximum (g) N=46 -2 0 2 4
Time from zero epoch (days)
SC24 - early declining (h) N=106
Figure 9.
Same as Figure 7 for the solar wind density. as was observed by
Jian et al. [2006] when studying SC23. The lowest values are observedin the late declining phase of SC23, when the peak values of the median curve just reach7 nT. These values are clearly smaller than in the late declining phase of the previous cycle,SC22. However, the difference comes mainly from the fact that the magnetic field valuesprior to the arrival of the SIR are clearly smaller in SC23, namely about 3 nT on day − B at 1 AU nearthe ecliptic plane during the 2007–2008 solar minimum period of SC23 have been a topic ofmany studies [see, e.g., Lee et al. , 2009;
Wu et al. , 2013].When comparing SC23 and SC24, a clear overall difference can be found. It appearsthat the SC23 data (panels b, c, and d) have their peaks of median B − . Again, the late declining phase ofSC23 is anomalous, since the background densities before the arrival of SIRs are only about3.5 cm − in SC23, while in SC22 the values are about 6 cm − . In the region of high speeds,the solar wind is dilute, and the densities are about 2.5 cm − in SC23, while in SC22 thedensities are about 4 cm − . Lee et al. [2009] noticed that when all solar wind OMNI data areconsidered, the peak occurrence for SC22 is centered at 3.5 cm − and for SC23 at 2.5 cm − .Obviously the rarefaction regions of HSSs dominate in the overall OMNI data.Figure 10 shows the superposed epoch analysis of Akasofu ε (expressed in GWh),which has been calculated by using equation (1). Since the ε parameter depends linearly onsolar wind velocity V and on the quadratic value of B , we expect that the cycle phases wherethese parameters maximize will have the largest values for ε . In addition, there is a strongdependence on the IMF clock angle, which is not shown in the previous plots. However,across several tens of events, the clock angle can be expected to take very different values,as the IMF will be northward in some events, southward in others, and with B z (cid:39) –15–onfidential manuscript submitted to JGR-Space Physics -2 0 2 4
Time from zero epoch (days)
SC22 - late declining (a) I med =12627 -2 0 2 4 Time from zero epoch (days)
SC23 - rising (b) I med =10928 -2 0 2 4 Time from zero epoch (days)
SC23 - maximum (c) I med =10013 -2 0 2 4 Time from zero epoch (days)
SC23 - early declining (d) I med =14028 -2 0 2 4 Time from zero epoch (days)
SC23 - late declining (e) I med =6181 -2 0 2 4 Time from zero epoch (days)
SC24 - rising (f) I med =5315 -2 0 2 4 Time from zero epoch (days)
SC24 - maximum (g) I med =7279 -2 0 2 4 Time from zero epoch (days)
SC24 - early declining (h) I med =10652 Figure 10.
Same as Figure 7 for the Akasofu ε parameter, except that the number given in the top-right-hand corner of each panel corresponds to the integrated value of the median curve from the zero epoch untilday 5 (in GWh). some other events. Therefore, one may expect that, in the superposed epoch analysis, theeffect of the clock angle on the value of Akasofu ε is averaged out. It can be seen that thehighest individual median value of the ε parameter takes place in the early declining phaseof SC23, where B has the largest values and V is relatively large. Almost as large values areseen in the late declining phase of SC22, which may be explained by the fact that the solarwind speed reached high values while B was still peaking.In order to assess the sustained impact of the SIR/HSSs on the geospace environment,the median values of the Akasofu ε parameter were also integrated from the zero epoch untilday 5, and those values are given in the panels of Figure 10 as I med . The early decliningphase of SC23 shows the largest integrated value of Akasofu ε . It also has the largest upperquartile values. The second highest integrated value is in the late declining phase of SC22,and the third highest value in the early declining phase of SC24. Even though the late declin-ing phase of SC23 had the largest velocities, the Akasofu ε parameter has low values, dueto the low values of B during the extended solar minimum, as discussed above. These arethe second lowest values of all the studied phases. The lowest values are found in the risingphase of SC24.Overall, Akasofu ε remains significantly lower during SC24 compared to SC23. Ineach phase, the energy input peaks around day 1 following the zero epoch, the value remainshigh until the end of day 2 and then decreases but remains somewhat elevated at least untilday 5.Figures 11 and 12 show the AE and SYM–H indices, respectively. In general, the max-imum of substorm ( AE ) and storm ( SYM–H ) activity is reached about one day after the zeroepoch, in the beginning of day 2, which is before the solar wind speed reaches its maximumand hence corresponds to the stream interaction region rather than the HSS itself.
Burlagaand Lepping [1977] showed that SIRs often amplify Alfvén waves propagating away fromthe Sun, which cause the IMF B z component to alternate between northward and south- –16–onfidential manuscript submitted to JGR-Space Physics -2 0 2 4
Time from zero epoch (days) AE ( n T ) SC22 - late declining (a) I med =24562 -2 0 2 4 Time from zero epoch (days)
SC23 - rising (b) I med =20576 -2 0 2 4 Time from zero epoch (days)
SC23 - maximum (c) I med =18105 -2 0 2 4 Time from zero epoch (days)
SC23 - early declining (d) I med =26537 -2 0 2 4 Time from zero epoch (days) AE ( n T ) SC23 - late declining (e) I med =15668 -2 0 2 4 Time from zero epoch (days)
SC24 - rising (f) I med =11508 -2 0 2 4 Time from zero epoch (days)
SC24 - maximum (g) I med =15166 -2 0 2 4 Time from zero epoch (days)
SC24 - early declining (h) I med =20951 Figure 11.
Same as Figure 10 for the AE index (integrated values in nTh). -2 0 2 4 Time from zero epoch (days) -40-30-20-1001020 SY M - H ( n T ) SC22 - late declining (a) I med =-2083 -2 0 2 4 Time from zero epoch (days) -40-30-20-1001020
SC23 - rising (b) I med =-1750 -2 0 2 4 Time from zero epoch (days) -40-30-20-1001020
SC23 - maximum (c) I med =-1386 -2 0 2 4 Time from zero epoch (days) -40-30-20-1001020
SC23 - early declining (d) I med =-1871 -2 0 2 4 Time from zero epoch (days) -40-30-20-1001020 SY M - H ( n T ) SC23 - late declining (e) I med =-1422 -2 0 2 4 Time from zero epoch (days) -40-30-20-1001020
SC24 - rising (f) I med =-1152 -2 0 2 4 Time from zero epoch (days) -40-30-20-1001020
SC24 - maximum (g) I med =-1333 -2 0 2 4 Time from zero epoch (days) -40-30-20-1001020
SC24 - early declining (h) I med =-1493 Figure 12.
Same as Figure 10 for the
SYM–H index (integrated values in nTh).–17–onfidential manuscript submitted to
JGR-Space Physics ward and therefore produce intermittent substorm activity while the SIR passes by Earth.Alfvén waves are also often present throughout the fast streams, which are also geoeffective.This is illustrated by the fact that the AE index remains enhanced at least until day 5. It iswell-known that the geoeffectiveness of HSSs in terms of substorm activity is also largelycontrolled by the sign of B z [e.g. Kavanagh et al. , 2012], which explains the rather largedispersion of the AE index values following the zero epoch indicated by the upper and lowerquartile values in blue and the fairly broad AE peak.Akasofu ε orders the AE index values quite well, so that the three most intense phasesand the weakest one (rising phase of SC24) are the same for AE as for Akasofu ε . The sameholds for SYM–H , except that the order of the two most intense events is reversed. Typically,the storms produced by SIR/HSSs are rather weak and only a few percent of the SIRs canproduce intense geomagnetic storms (
Dst < −
100 nT) [
Richardson et al. , 2006]. However,HSSs are quite effective in producing substorms, and these events are called HILDCAA(high-intensity long-duration auroral activity) by
Tsurutani et al. [2006]. It was found by
Hajra et al. [2013] that HILDCAAs are on average 20% longer during the declining phasethan during the rising and maximum phases. Figure 11 shows indeed that during the declin-ing phases the AE index is still elevated on day 5. In addition, the maximum AE values arehigher during the early declining phases of SC23 and SC24 than during the other phases.Again, the top-row panels of Figures 11 and 12 exhibit a more intense response thanthe bottom-row panels, underlining that geomagnetic activity related to SIR/HSSs signifi-cantly decreased between the early and late declining phases of SC23. Within a given solarcycle, the rising and maximum phases are those with lowest geomagnetic activity duringSIR/HSS events, whereas the early and late declining phases exhibit stronger activity.Figure 13 visualizes the 5-day integrated values of geomagnetic activity measured by AE (orange bars) and SY M – H (blue bars) during SIR/HSS events within the defined solarcycle phases from 1995 to 2017, superposed on the monthly sunspot number (black line).This figure shows the well-known importance of declining phases of the solar cycles forgeomagnetic activity. For SC22, we have data only for the late declining phase, and it is themost geoeffective one in terms of SY M – H , and the second-most geoeffective in terms of AE for all data. Inside SC23, the early declining phase is the most geoeffective one in terms ofboth indices, and in terms of AE the most geoeffective one of all data. The anomalous latedeclining phase of SC23 is the least geoeffective one, and even the rising phase of this cycleis more geoeffective. The maximum phase is less geoeffective than the rising phase in SC23.When moving to SC24, the geoeffectiveness clearly decreases compared to SC23, as doesthe sunspot number. The early declining phase is the most geoeffective one inside SC24, butwe do not yet have data for the late declining phase. Contrary to SC23, the maximum phaseis more geoeffective than the rising phase. Comparing SC24 to SC23 in terms of auroralelectrojet and substorm activity measured by the AE index, the geoeffectiveness has beenabout 40% lower in the rising phase, 15% lower in the maximum phase, and 20% lower inthe early declining phase. For the ring current and storms measured by the SY M – H index,the geoeffectiveness has been about 35% lower in the rising phase and 20% lower in the earlydeclining phase, but roughly similar during the maximum phase.As a summary, we find that the geoeffectiveness of SIR/HSSs varies in the differentphases of the solar cycle. Our results are in line with the often presented statement that SIR/HSSsare the most geoeffective during the declining phases of solar cycle [e.g., Hajra et al. , 2013;
Kilpua et al. , 2017;
Chi et al. , 2018], even though we find that the late declining phase ofSC23 was very weakly geoeffective owing to reasons discussed in the next section.
We will discuss two major topics. First, the validity of the detection algorithm is as-sessed by comparing the obtained list of SIR/HSS events with existing catalogs during years –18–onfidential manuscript submitted to
JGR-Space Physics
Year -2000-1500-1000-5000 I m ed ( SY M - H ) [ n T h ] S C S C S C I m ed ( AE ) [ n T h ] S un s po t nu m be r Figure 13.
Bar diagram: Integrated values of median curves from the zero epoch until day 5 for the AE (top) and SY M – H (bottom) indices (see Figures 11 and 12). Line plot: Monthly sunspot number.–19–onfidential manuscript submitted to JGR-Space Physics when they overlap. Second, the identified solar cycle phase-dependent features of SIR/HSSsand their geoeffectiveness will be considered in the global picture of the known features ofthe studied solar cycles.
We compared the events detected by our algorithm with the HSS catalog issued byMaris , Muntean et al. ( ), which covers years from 2009until 2016. We selected year 2015 for the comparison, as it is at the beginning of the de-clining phase of SC24 and hence contains a mix of HSSs, ICMEs and slow solar wind. TheMaris , Muntean et al. catalog is solely based on the solar wind speed, averaged at 3 h reso-lution. A HSS event is detected if the maximum 3 h mean speed during one day exceeds theminimum 3 h mean speed of the previous day by at least 100 km/s, and if the enhanced speedpersists for two days. In 2015, with their method, Maris , Muntean et al. detected 44 HSSevents. Comparatively, our algorithm detected only 27 SIR/HSS events during that year.Two main reasons explain this first difference: (i) the Maris , Muntean et al. catalog does notfocus on SIR-related HSSs, and hence contains 11 events corresponding to (or contaminatedby) ICMEs (by comparison with the Richardson and Cane list), and (ii) the Maris , Munteancatalog is not restricted to HSSs with V max ≥
500 km/s, and therefore contains another10 events whose maximum velocities lie between 400 and 500 km/s. On the other hand, ouralgorithm detected four SIR/HSS events which are not in the Maris , Muntean et al. catalog(on 14 July, 18 August, 2 September, and 28 November 2015). Figure S1 in SupportingInformation illustrates the comparison of the two lists for 2015. For the 23 events common toboth catalogs, 16 of them have their starting time with less than 6 h difference, and only oneof them has starting time differing by more than 12 hours (2 January, 23 UT vs 4 January, 9–12 UT). For this specific event, our starting date nicely corresponds to the solar wind densitypeak, which makes this detection time consistent with the rest of the list (see Figure 9).Another HSS catalog was earlier compiled by
Gupta and Badruddin [2010] for years1996–2007. The retained criteria are essentially the same as those used by Maris , Munteanet al., i.e., an increase in the solar wind speed of at least 100 km/s with sustained veloc-ity for at least two days. Year 1996 was chosen for comparison of our list with the
Guptaand Badruddin [2010] catalog, since it corresponds to the peak of SIR/HSS activity at thetransition between SC22 and SC23.
Gupta and Badruddin [2010] listed 40 HSS events in1996, but several of them actually contained two to three distinct streams, therefore repre-senting at total of 58 individual streams (compared to 35 events from our method). Amongthese, 22 events were associated with solar wind speeds remaining below our threshold of500 km/s. One of their events (19 September) was not detected by our algorithm, since itsinitial solar wind speed was greater than 450 km/s (see criterion 2a), and another one (thesecond stream of the event starting on 8 April) was also rejected for being too soon afterour 11 April event (see criterion 3). On the other hand, our algorithm detected a SIR/HSSevent on 2 July, which is not included in the
Gupta and Badruddin [2010] catalog. Besidesthese few differences, there is an overall good agreement between both catalogs. Figure S2 inSupporting Information illustrates the comparison of the two lists for 1996.A third list of HSSs which may be used for comparison is the one given in
Morleyet al. [2010], which uses different criteria for SIR/HSS identification. As a first criterion,
Morley et al. [2010] looked for an enhancement in the radial velocity of the solar wind ( V x )associated with a west–east deflection of the flow, i.e., a change in the sign of V y . For theobtained SIR/HSS candidates, other solar wind parameters such as the IMF magnitude, theproton number density and the proton temperature were then examined to reject false posi-tives. In addition, two consecutive events had to be separated by at least 2 days, and eventsexhibiting clear ICME signatures were removed from the list. The list given in Table 2 of Morley et al. [2010] considers 67 stream interfaces verifying those criteria between 2005 and2008. During these same years, our algorithm detected 136 SIR/HSS events, which is morethan twice as many. Yet, of the 67 events studied by
Morley et al. [2010], we found that nine –20–onfidential manuscript submitted to
JGR-Space Physics of them were not detected by our algorithm. In two of these cases, a ICME was reported inthe Richardson and Cane list (criterion 4), in three cases, the maximum solar wind speed didnot reach 500 km/s (criterion 2c), and in four cases a previous event was detected less thanthree days before the rejected event (criterion 3). Figures S3–S6 in Supporting Informationillustrate the comparison of the two lists for 2005–2008.For the 58 SIR/HSS events in common with the
Morley et al. [2010] list, a comparisonof the starting times reveals that in all cases but one our algorithm gives an earlier time. Thisis inherent to the difference in the criteria for setting the beginning of the events: while ouralgorithm retains the time after which the solar wind speed starts increasing,
Morley et al. [2010] were interested in the stream interface itself, which is determined with the azimuthalflow reversal ( V y sign change). The difference ∆ t between those starting times can varysignificantly. Figure S7 in Supporting Information gives a histogram of ∆ t , which is definedas positive when our algorithm gives an earlier time than in the Morley et al. [2010] list.The peak in the distribution of ∆ t is for values comprised between 6 and 12 hours, but upto 70 hour difference was found for a couple of events. The first such event (3 May 2006)exhibits steady solar wind speed enhancement for three days before the stream interface,and in the second case (7 February 2008), our method detects a small stream immediatelypreceding the main event, hence causing an erroneously early starting time. The differenceis negative in only one case (22 April 2008), for which the start time given by our algorithmcorresponds nicely to the beginning of the solar wind speed increase.Finally, the list obtained with the detection method presented in this paper can be com-pared to the catalog compiled by Jian et al. [2006], later extended to cover years 1995 to2009 (extended version available at ). Between 1995 and 2009, the
Jian et al. [2006]list contains 437 SIR/HSSs with V max ≥
500 km/s, whereas our list contains 394 eventsduring the same time 15-year period. When making a detailed analysis of the differencesbetween those two lists, it appears that 39 events from our list are not present in the Jian etal. list. These events have each been examined visually to ensure they do exhibit typicalSIR/HSS signatures. The reason for the difference may be that the event identification by
Jian et al. [2006] imposes that the events must exhibit at least five out of the seven expectedfeatures considered in their study, which are solar wind speed increase, total perpendic-ular pressure pileup, velocity deflections, proton density enhancement, proton tempera-ture enhancement, increase in entropy, and magnetic field compression. On the other hand,82 events with V max ≥
500 km/s in the Jian et al. list are not present in ours. The reasonsare: presence of an ICME with V max ≥
500 km/s within [ t − t + V max <
500 km/s in the hourly OMNI data whereas the Wind or ACE data with 93 s and 64 sresolution, respectively, used by Jian et al. exceeds 500 km/s during the event (25 events);solar wind speed greater than 450 km/s at the start time of the event (15 events); compoundevents separated by less than 3 days (9 events); no sharp IMF magnitude gradient (1 event).Besides, in 27 cases, event starting dates differed by more than 24 h from one list to theother. About half of these cases correspond to compound events for which in one list thefirst event only is mentioned, while in the other list the second event only is mentioned. Inthe other cases, our list gives an earlier time than the Jian et al. list because the detected IMFmagnitude enhancement is earlier than the leading edge of the SIR as identified by Jian et al.Overall, the agreement between both lists over those 15 years of comparison is satisfactory,and all the differences can be explained by the choices made in the criteria for identificationin this study versus the Jian et al. study.While some differences may be noted with each of those above-discussed catalogs,one advantage of the list obtained with this algorithm is that it contains SIR/HSS eventsduring 23 years, i.e., roughly two solar cycles. Provided an updated list of ICMEs reachingthe Earth exists and solar wind observations remain available, this list of SIR/HSS eventscan be completed during the upcoming years by applying the same algorithm, thus enablingfurther studies about the upcoming solar cycles. –21–onfidential manuscript submitted to
JGR-Space Physics
We have briefly discussed in the previous sections the properties of the solar windat 1 AU near the ecliptic plane during the extended solar minimum of SC23 observed inother studies. Comparing the minimum periods of SC22 and SC23,
Lee et al. [2009] noticedthe following. For the IMF magnitude, the peak of the distribution for the SC23 period iscentered at 3.5 nT, which is 30% less than 5 nT, the approximate central value for the peak ofthe SC22 distribution. For the number density, the distribution is shifted toward lower valuesby about 30%. For solar wind velocity, both the SC22 and SC23 velocity distributions havepeak values occurring around 340 km/s, but for SC23 the peak is lower and the high-speedtail distribution, which is centered near 580 km/s, has slightly larger percent occurrenceduring SC23.In this study, we have specifically studied the SIR/HSS events and our findings arein general accordance with
Lee et al. [2009], but give some additional insight into the dis-tribution functions. We found by comparing the late declining phases of SC22 and SC23,namely years 1995–1996 and 2006–2008, the following properties. Both the IMF magnitudeand density values before the SIR/HSS events are about 30% lower in SC23 than SC22,but the compressions of B and N are equally strong in both cycles. The peak of the densitydistribution in Lee et al. [2009] study comes obviously from the dilute solar wind in the high-speed stream region (in our figures superposed epoch after day 2).Comparison of the distributions of HSS peak velocities during the late declining phasesof SC22 and SC23 shown in Figure 6 indicates that the proportion of 500 < V max <
600 km/sis higher in SC22, whereas the proportion of V max >
700 km/s is higher in SC23. Thesuperposed epoch presentation of velocities shown in Figure 7 confirms that the median peakvelocities are higher in SC23 ( V max ∼
560 km/s) than SC22 ( V max ∼
510 km/s). These arein harmony with the observations of the behavior of the high-speed tail distributions by
Leeet al. [2009].The solar wind parameters control the coupling of the IMF with the terrestrial magne-tosphere and energy transfer from the solar wind into the magnetosphere–ionosphere system.Several coupling functions have been developed [see, e.g.,
Newell et al. , 2007], which alldepend on solar wind velocity and magnetic field. In this study, we have used the traditionalAkasofu ε . We saw a very similar behavior between the Akasofu ε and the AE and SY M – H indices. This can be expected, since it is well-known that the solar wind parameters di-rectly affect the geoeffectiveness of SIR/HSSs. The surprisingly poor geoeffectiveness of theSIR/HSS events in the late declining phase of SC23 is obviously explained by the unusuallylow magnetic field values during the deep minimum of SC23 discussed above. Even thoughthe velocities of these HSSs were higher than in the late declining phase of SC22, this did notmake those more geoeffective, and this can be explained by the quadratic dependence on themagnetic field intensity compared to the linear dependence on the velocity in the couplingfunction. The fact that SIR/HSSs are less geoeffective during SC24 than during SC23 alsofinds an explanation in the solar wind properties. The velocities were a bit smaller duringSC24 than SC23, but the magnetic fields were even more significantly smaller. This will bediscussed more below.Following the extremely low minimum of SC23, SC24 has shown unusually low activ-ity [ Kamide and Kusano , 2013]. In particular,
McComas et al. [2013] showed that the risingphase of SC24 exhibited significantly lower key parameters (e.g., IMF magnitude, solar windproton density, temperature, velocity, dynamic pressure) than the rising phases of previoussolar cycles, which correlates with the fact that SC24 has been the weakest solar cycle sincethe beginning of the Space Age. Figures 7–12 indicate that this conclusion hold not onlywhen looking at solar activity as a whole but also when focusing specifically on solar windconditions during SIR/HSSs. It is noteworthy that, while the IMF magnitude (see Figure 8)had very low values during the late declining phase of SC23 as discussed above, the HSSpeak velocity was meanwhile at its highest values, and started to drop only during the rising –22–onfidential manuscript submitted to
JGR-Space Physics phase of SC24 (see Figure 7) and remained at lower values than during the previous cycle.Even though the magnetic field has to some extent recovered from the deep minimum ofSC23, it has continued to be at a lower level during SC24 than at the corresponding phasesduring SC23.In fact, the early/late declining phase distinction for SC24 cannot presently be deter-mined, as the minimum is still to be reached. Whether such a distinction will even be neededfor SC24 is not certain, as it may be so that the unusually long declining phase of SC23 andthe associated extremely weak activity conditions were unique.
We presented a method for detection of solar wind high-speed stream events, based onthe IMF magnitude time derivative, the solar wind velocity, and, to remove ICMEs misiden-tified as SIR/HSSs and SIR/HSSs containing an embedded ICME, comparison with theRichardson and Cane list of ICMEs. This algorithm has been applied to years 1995–2017,leading to the detection of 588 SIR/HSS events with speeds exceeding 500 km/s. Whencompared to the existing HSS lists, which use slightly different criteria and may includea human visual detection of events, the algorithm proves efficient in correctly identifyingSIR/HSS events and setting their beginning at the time when the solar wind speed startsincreasing. The main benefit of this new detection method is that it produces a catalog ofSIR/HSS events over 23 years of solar wind observations in a transparent and reproducibleway. The analysis of the 588 SIR/HSS events obtained with the detection method shows thatthe yearly number of SIR/HSS events peaks during the declining phase of the solar cycle.Besides the maximum yearly number of SIR/HSS events in SC23 (39 events in 2007) isabout the same as the maximum number in 1995 of SC22 (35 events) and in 2016 of SC24(35 events). It is yet to be seen if the yearly number of SIR/HSS events in SC24 increaseswhen approaching the solar cycle minimum.The number of SIR/HSS events sharply drops after the solar minimum is reached,during the first year of a new cycle. After that, the yearly number of events starts to rise.Even during the rising phases, which are associated with the smallest number of SIR/HSSs,the yearly number of SIR/HSSs amounts to 69% and 48% of the yearly number during thedeclining phases of SC23 and SC24, respectively. During the solar maximum years, whichare dominated by ICME events, we find a local minimum in the occurrence of SIR/HSSevents.During the late declining phase of SC23, large low-latitude coronal holes persistedon the solar surface, which had long lifetimes. This resulted in repeated very-high-velocitystreams. Even though the compressions of magnetic field were of the same magnitude asduring the late declining phase of the previous SC22, the low background magnetic field val-ues at 1 AU near the ecliptic plane during the extended minimum of 2007–2008 [consistentwith results by
Lee et al. , 2009;
Wu et al. , 2013] resulted in low B values in the ecliptic SIRregions and in low solar wind coupling function (Akasofu ε ) values. Hence, the geoeffec-tiveness of SIR/HSSs during the late declining phase of SC23 was the smallest of this entirecycle.During SC24, both the solar wind velocity and the magnetic field have continued to beat a lower level than during the corresponding phases of SC23. Hence, the geoeffectivenessof SIR/HSSs has been lower too. For auroral electrojet and substorm activity measured bythe AE index, the geoeffectiveness has been about 40% lower during the rising phase and20% lower during the early declining phase. For the ring current and storms measured by the SY M – H index, the geoeffectiveness has been about 35% lower during the rising phase and20% lower during the early declining phase. –23–onfidential manuscript submitted to JGR-Space Physics
We can confirm the earlier-reported fact the SIR/HSS events have highest geoeffec-tiveness during the declining phases. However, there may be solar-cycle-dependent dif-ferences when one considers the early and late declining phases. The late declining phaseof SC22 was very geoeffective, whereas the corresponding phase in SC23 was not. Theearly declining phase was the most geoeffective one inside SC23. So far, for SC24, the earlydeclining phase has also been the most geoeffective one; however, we do not know yet howgeoeffective SIR/HSSs will be during the late declining phase of SC24.
Acknowledgments
This work is supported by the Academy of Finland (projects 312351 and 285474) and theEuropean Research Council (Consolidator Grant 682068-PRESTISSIMO). The OMNI datawere obtained from the GSFC/SPDF OMNIWeb interface at http://omniweb.gsfc.nasa.gov . The monthly values of the sunspot number between 1995 and 2017 were re-trieved from the World Data Center SILSO, Royal Observatory of Belgium, Brussels. TheRichardson and Cane list of ICMEs was downloaded from . The solar cycle beginning dates weredetermined from ftp://ftp.ngdc.noaa.gov/STP/space-weather/solar-data/solar-indices/sunspot-numbers/cycle-data/table_cycle-dates_maximum-minimum.txt . This work utilizes the HSS catalog issued by G. Maris , Muntean, D. Bes , liu–Ionescu andV. Dobrică, managed by the Institute of Geodynamics of the Romanian Academy. References
Akasofu, S.-I. (1979), A search for the interplanetary quantity controlling the development ofgeomagnetic storms,
Royal Astronomical Society, Quarterly Journal , , 119–137.Al-Shakarchi, D. A., and H. Morgan (2018), Properties of the hps-icme-cir interactionevent of 9–10 september 2011, J. Geophys. Res. Space Physics , (4), 2535–2556, doi:10.1002/2017JA024849.Alves, M. V., E. Echer, and W. D. Gonzalez (2006), Geoeffectiveness of corotating inter-action regions as measured by Dst index, J. Geophys. Res. Space Physics , , A07S05,doi:10.1029/2005JA011379.Bame, S. J., J. R. Asbridge, W. C. Feldman, and J. T. Gosling (1976), Solar cycle evolutionof high-speed solar wind streams., The Astrophysical Journal , , 977–980, doi:10.1086/154566.Belcher, J. W., and J. Davis, Leverett (1971), Large-amplitude Alfvén waves in theinterplanetary medium, 2, Journal of Geophysical Research , , 3534, doi:10.1029/JA076i016p03534.Borovsky, J. E., and M. H. Denton (2006), Differences between CME-driven stormsand CIR-driven storms, J. Geophys. Res. Space Physics , , A07S08, doi:10.1029/2005JA011447.Borovsky, J. E., and M. H. Denton (2009), Relativistic-electron dropouts and recovery: Asuperposed epoch study of the magnetosphere and the solar wind, J. Geophys. Res. SpacePhysics , , A02201, doi:10.1029/2008JA013128.Burlaga, L. F., and R. P. Lepping (1977), The causes of recurrent geomagnetic storms, Planetary and Space Science , , 1151–1160, doi:10.1016/0032-0633(77)90090-3.Chi, Y., C. Shen, B. Luo, Y. Wang, and M. Xu (2018), Geoeffectiveness of Stream In-teraction Regions From 1995 to 2016, Space Weather , , 1960–1971, doi:10.1029/2018SW001894.Chkhetiia, A. M. (1975), Global distribution of the 27-day electron density variation at themaximum of the ionospheric F layer, Geomagnetism and Aeronomy , , 160.Chree, C. (1913), Some phenomena of sunspots and of terrestrial magnetism at KewObservatory, Philosophical Transactions of the Royal Society of London Series A , ,75–116, doi:10.1098/rsta.1913.0003. –24–onfidential manuscript submitted to JGR-Space Physics
Crowley, G., A. Reynolds, J. P. Thayer, J. Lei, L. J. Paxton, A. B. Christensen, Y. Zhang,R. R. Meier, and D. J. Strickland (2008), Periodic modulations in thermosphericcomposition by solar wind high speed streams,
Geophysical Research Letters , , L21106,doi:10.1029/2008GL035745.Davis, T. N., and M. Sugiura (1966), Auroral electrojet activity index AE and its universaltime variations, J. Geophys. Res. , , 785–801, doi:10.1029/JZ071i003p00785.de Toma, G. (2011), Evolution of coronal holes and implications for high-speed solar windduring the minimum between cycles 23 and 24, Solar Physics , (1-2), 195–217.Denton, M. H., and J. E. Borovsky (2012), Magnetosphere response to high-speed solarwind streams: A comparison of weak and strong driving and the importance of extendedperiods of fast solar wind, J. Geophys. Res. Space Physics , , A00L05, doi:10.1029/2011JA017124.Denton, M. H., T. Ulich, and E. Turunen (2009), Modification of midlatitude ionosphericparameters in the F2 layer by persistent high-speed solar wind streams, Space Weather , ,S04006, doi:10.1029/2008SW000443.Echer, E., B. T. Tsurutani, and W. D. Gonzalez (2013), Interplanetary origins of moderate( −
100 nT < Dst ≤ −
50 nT) geomagnetic storms during solar cycle 23 (1996–2008),
J.Geophys. Res. Space Physics , (1), 385–392, doi:10.1029/2012JA018086.Gonzalez, W. D., B. T. Tsurutani, and A. L. Clúa de Gonzalez (1999), Interplanetaryorigin of geomagnetic storms, Space Science Reviews , , 529–562, doi:10.1023/A:1005160129098.Gopalswamy, N., and M. Kundu (1992), Estimation of the mass of a coronal mass ejectionfrom radio observations, The Astrophysical Journal , , L37–L39.Gosling, J. T., and V. J. Pizzo (1999), Formation and Evolution of Corotating InteractionRegions and their Three Dimensional Structure, Space Science Reviews , , 21–52, doi:10.1023/A:1005291711900.Gosling, J. T., A. J. Hundhausen, V. Pizzo, and J. R. Asbridge (1972), Compressions andrarefactions in the solar wind: Vela 3, Journal of Geophysical Research , (28), 5442–5454, doi:10.1029/JA077i028p05442.Gosling, J. T., J. R. Asbridge, S. J. Bame, and W. C. Feldman (1976), Solar wind speedvariations: 1962-1974, J. Geophys. Res. , , 5061–5070, doi:10.1029/JA081i028p05061.Grandin, M. (2017), Multi-instrument and modelling studies of the iono-spheres at Earth and Mars , (Doctoral dissertation). Retrieved from Jultika(http://urn.fi/urn:isbn:9789526216157). Sodankylä (Finland): University of Oulu.Grandin, M., A. T. Aikio, A. Kozlovsky, T. Ulich, and T. Raita (2015), Effects of solar windhigh-speed streams on the high-latitude ionosphere: Superposed epoch study,
J. Geophys.Res. Space Physics , , 10,669–10,687, doi:10.1002/2015JA021785.Grandin, M., A. T. Aikio, A. Kozlovsky, T. Ulich, and T. Raita (2017), Cosmic radio noiseabsorption in the high-latitude ionosphere during solar wind high-speed streams, J.Geophys. Res. Space Physics , , 5203–5223, doi:10.1002/2017JA023923.Gupta, V., and Badruddin (2010), High-Speed Solar Wind Streams during 1996–2007:Sources, Statistical Distribution, and Plasma/Field Properties, Solar Physics , , 165–188, doi:10.1007/s11207-010-9554-z.Hajra, R., E. Echer, B. T. Tsurutani, and W. D. Gonzalez (2013), Solar cycle dependence ofHigh-Intensity Long-Duration Continuous AE Activity (HILDCAA) events, relativisticelectron predictors?, Journal of Geophysical Research (Space Physics) , , 5626–5638,doi:10.1002/jgra.50530.Hapgood, M. A. (1993), A double solar cycle in the 27-day recurrence of geomagneticactivity, Annales Geophysicae , , 248–253.Heidrich-Meisner, V., T. Peleikis, M. Kruse, L. Berger, and R. F. Wimmer-Schweingruber(2017), Evolution of an equatorial coronal hole structure and the released coronal holewind stream: Carrington rotations 2039 to 2050, Astronomy & Astrophysics , , A84,doi:10.1051/0004-6361/201730574. –25–onfidential manuscript submitted to JGR-Space Physics
Iyemori, T. (1990), Storm-time magnetospheric currents inferred from mid-latitudegeomagnetic field variations,
Journal of Geomagnetism and Geoelectricity , , 1249–1265.Jian, L., C. T. Russell, J. G. Luhmann, and R. M. Skoug (2006), Properties of StreamInteractions at One AU During 1995 2004, Solar Physics , , 337–392, doi:10.1007/s11207-006-0132-3.Jian, L. K., C. T. Russell, and J. G. Luhmann (2011), Comparing Solar Minimum 23/24with Historical Solar Wind Records at 1 AU, Solar Physics , , 321–344, doi:10.1007/s11207-011-9737-2.Kamide, Y., and K. Kusano (2013), Is Something Wrong With the Present Solar Maximum?, Space Weather , , 140–141, doi:10.1002/swe.20045.Kavanagh, A. J., F. Honary, E. F. Donovan, T. Ulich, and M. H. Denton (2012), Keyfeatures of >
30 keV electron precipitation during high speed solar wind streams: Asuperposed epoch analysis,
J. Geophys. Res. Space Physics , , A00L09, doi:10.1029/2011JA017320.Kilpua, E. K. J., A. Balogh, R. von Steiger, and Y. D. Liu (2017), Geoeffective Properties ofSolar Transients and Stream Interaction Regions, Space Science Reviews , , 1271–1314,doi:10.1007/s11214-017-0411-3.King, J. H., and N. E. Papitashvili (2005), Solar wind spatial scales in and comparisons ofhourly Wind and ACE plasma and magnetic field data, J. Geophys. Res. Space Physics , , A02104, doi:10.1029/2004JA010649.Krieger, A. S., A. F. Timothy, and E. C. Roelof (1973), A Coronal Hole and Its Identificationas the Source of a High Velocity Solar Wind Stream, Solar Physics , , 505–525, doi:10.1007/BF00150828.Krista, L. D., S. W. McIntosh, and R. J. Leamon (2018), The Longitudinal Evolution ofEquatorial Coronal Holes, The Astronomical Journal , (4), 153.Lee, C., J. Luhmann, X. Zhao, Y. Liu, P. Riley, C. Arge, C. Russell, and I. de Pater (2009),Effects of the weak polar fields of solar cycle 23: Investigation using omni for the stereomission period, Solar physics , (1-2), 345–363.Lindsay, G. M., J. G. Luhmann, C. T. Russell, and J. T. Gosling (1999), Relationshipsbetween coronal mass ejection speeds from coronagraph images and interplanetarycharacteristics of associated interplanetary coronal mass ejections, Journal of GeophysicalResearch , , 12,515–12,524, doi:10.1029/1999JA900051.Mathie, R. A., and I. R. Mann (2001), On the solar wind control of Pc5 ULF pulsation powerat mid-latitudes: Implications for MeV electron acceleration in the outer radiation belt, J.Geophys. Res. , , 29,783–29,796, doi:10.1029/2001JA000002.McComas, D. J., S. J. Bame, P. Barker, W. C. Feldman, J. L. Phillips, P. Riley, and J. W.Griffee (1998), Solar Wind Electron Proton Alpha Monitor (SWEPAM) for the Ad-vanced Composition Explorer, Space Science Reviews , , 563–612, doi:10.1023/A:1005040232597.McComas, D. J., N. Angold, H. A. Elliott, G. Livadiotis, N. A. Schwadron, R. M. Skoug,and C. W. Smith (2013), Weakest solar wind of the space age and the current "mini" solarmaximum, The Astrophysical Journal , (1), 2.McPherron, R. L., and J. Weygand (2006), The Solar Wind and Geomagnetic Activity asa Function of Time Relative to Corotating Interaction Regions, in Recurrent MagneticStorms: Corotating Solar Wind , vol. 167, p. 125, doi:10.1029/167GM12.Meredith, N. P., R. B. Horne, M. M. Lam, M. H. Denton, J. E. Borovsky, and J. C. Green(2011), Energetic electron precipitation during high-speed solar wind stream drivenstorms,
J. Geophys. Res. Space Physics , , A05223, doi:10.1029/2010JA016293.Morley, S., R. Friedel, E. Spanswick, G. Reeves, J. Steinberg, J. Koller, T. Cayton, andE. Noveroske (2010), Dropouts of the outer electron radiation belt in response to solarwind stream interfaces: global positioning system observations, The Royal Society , (2123), 3329–3350, doi:10.1098/rspa.2010.0078. –26–onfidential manuscript submitted to JGR-Space Physics
Neugebauer, M., D. Reisenfeld, and I. G. Richardson (2016), Comparison of algorithms fordetermination of solar wind regimes,
Journal of Geophysical Research (Space Physics) , , 8215–8227, doi:10.1002/2016JA023142.Newell, P. T., T. Sotirelis, K. Liou, C. I. Meng, and F. J. Rich (2007), A nearly universal solarwind-magnetosphere coupling function inferred from 10 magnetospheric state variables, J.Geophys. Res. Space Physics , , A01206, doi:10.1029/2006JA012015.Ogilvie, K. W., D. J. Chornay, R. J. Fritzenreiter, F. Hunsaker, J. Keller, J. Lobell, G. Miller,J. D. Scudder, E. C. Sittler, Jr., R. B. Torbert, D. Bodet, G. Needell, A. J. Lazarus, J. T.Steinberg, J. H. Tappan, A. Mavretic, and E. Gergin (1995), SWE, A ComprehensivePlasma Instrument for the Wind Spacecraft, Space Science Reviews , , 55–77, doi:10.1007/BF00751326.Reisenfeld, D. B., R. C. Wiens, B. L. Barraclough, J. T. Steinberg, M. Neugebauer, J. Raines,and T. H. Zurbuchen (2013), Solar Wind Conditions and Composition During the GenesisMission as Measured by in situ Spacecraft, Space Science Reviews , , 125–164, doi:10.1007/s11214-013-9960-2.Richardson, I. G. (2018), Solar wind stream interaction regions throughout the heliosphere, Living Reviews in Solar Physics , , 1, doi:10.1007/s41116-017-0011-z.Richardson, I. G., and H. V. Cane (2010), Near-Earth Interplanetary Coronal Mass EjectionsDuring Solar Cycle 23 (1996 - 2009): Catalog and Summary of Properties, Solar Physics , , 189–237, doi:10.1007/s11207-010-9568-6.Richardson, I. G., and H. V. Cane (2012a), Solar wind drivers of geomagnetic storms duringmore than four solar cycles, Journal of Space Weather and Space Climate , , A01, doi:10.1051/swsc/2012001.Richardson, I. G., and H. V. Cane (2012b), Near-earth solar wind flows and related geomag-netic activity during more than four solar cycles (1963-2011), Journal of Space Weatherand Space Climate , , A02, doi:10.1051/swsc/2012003.Richardson, I. G., E. W. Cliver, and H. V. Cane (2000), Sources of geomagnetic activity overthe solar cycle: Relative importance of coronal mass ejections, high-speed streams, andslow solar wind, J. Geophys. Res. , , 18,203–18,213, doi:10.1029/1999JA000400.Richardson, I. G., E. W. Cliver, and H. V. Cane (2001), Sources of geomagnetic storms forsolar minimum and maximum conditions during 1972-2000, Geophys. Res. Lett. , ,2569–2572, doi:10.1029/2001GL013052.Richardson, I. G., D. F. Webb, J. Zhang, D. B. Berdichevsky, D. A. Biesecker, J. C. Kasper,R. Kataoka, J. T. Steinberg, B. J. Thompson, C. C. Wu, and A. N. Zhukov (2006), Majorgeomagnetic storms (Dst < -100 nT) generated by corotating interaction regions, J.Geophys. Res. Space Physics , , A07S09, doi:10.1029/2005JA011476.Richter, A. K., and A. H. Luttrell (1986), Superposed epoch analysis of corotating interactionregions at 0.3 and 1.0 au: A comparative study, J. Geophys. Res. Space Physics , (A5),5873–5878, doi:10.1029/JA091iA05p05873.Schwenn, R. (1983), Direct Correlations Between Coronal Transients and InterplanetaryDisturbances, Space Science Reviews , , 85–99, doi:10.1007/BF00221199.Sheeley, J., N. R., R. A. Howard, D. J. Michels, M. J. Koomen, R. Schwenn, K. H.Muehlhaeuser, and H. Rosenbauer (1985), Coronal mass ejections and interplanetaryshocks, Journal of Geophysical Research , , 163–176, doi:10.1029/JA090iA01p00163.Sheeley Jr, N., J. Harvey, and W. Feldman (1976), Coronal holes, solar wind streams, andrecurrent geomagnetic disturbances: 1973–1976, Solar Physics , (2), 271–278.Shugay, Y., V. Slemzin, D. Rodkin, Y. Yermolaev, and I. Veselovsky (2018), Influence ofcoronal mass ejections on parameters of high-speed solar wind: a case study, Journal ofSpace Weather and Space Climate , , A28.SILSO World Data Center (1995-2017), The International Sunspot Number.Smith, C. W., J. L’Heureux, N. F. Ness, M. H. Acuña, L. F. Burlaga, and J. Scheifele (1998),The ACE magnetic fields experiment, Space Science Reviews , , 613–632, doi:10.1023/A:1005092216668. –27–onfidential manuscript submitted to JGR-Space Physics
Smith, E. J., and J. H. Wolfe (1976), Observations of interaction regions and corotatingshocks between one and five au: Pioneers 10 and 11,
Geophysical Research Letters , (3),137–140, doi:10.1029/GL003i003p00137.Snyder, C. W., M. Neugebauer, and U. R. Rao (1963), The Solar Wind Velocity and ItsCorrelation with Cosmic-Ray Variations and with Solar and Geomagnetic Activity, Journal of Geophysical Research , , 6361, doi:10.1029/JZ068i024p06361.Temmer, M., B. Vršnak, and A. M. Veronig (2007), Periodic Appearance of Coronal Holesand the Related Variation of Solar Wind Parameters, Solar Physics , , 371–383, doi:10.1007/s11207-007-0336-1.Tsurutani, B. T., W. D. Gonzalez, A. L. C. Gonzalez, F. L. Guarnieri, N. Gopalswamy,M. Grande, Y. Kamide, Y. Kasahara, G. Lu, I. Mann, R. McPherron, F. Soraas, andV. Vasyliunas (2006), Corotating solar wind streams and recurrent geomagnetic activity:A review, J. Geophys. Res. Space Physics , , A07S01, doi:10.1029/2005JA011273.Wanliss, J. A., and K. M. Showalter (2006), High-resolution global storm index: Dst versusSYM-H, J. Geophys. Res. Space Physics , , A02202, doi:10.1029/2005JA011034.Webb, D. F., E. W. Cliver, N. U. Crooker, O. C. S. Cry, and B. J. Thompson (2000),Relationship of halo coronal mass ejections, magnetic clouds, and magnetic storms, Journal of Geophysical Research , , 7491–7508, doi:10.1029/1999JA000275.Wu, C.-C., K. Liou, S. Plunkett, C. D. Fry, and S.-T. Wu (2013), Investigation of so-lar/heliospheric anomalies associated with the solar minimum of 2007–2008, Terrestrial,Atmospheric & Oceanic Sciences , (2), 243–252, doi:10.3319/TAO.2012.10.16.01(SEC).Xu, F., and J. E. Borovsky (2015), A new four-plasma categorization scheme for the solarwind, J. Geophys. Res. Space Physics , , 70–100, doi:10.1002/2014JA020412.Zhang, J., I. G. Richardson, D. F. Webb, N. Gopalswamy, E. Huttunen, J. C. Kasper, N. V.Nitta, W. Poomvises, B. J. Thompson, C. C. Wu, S. Yashiro, and A. N. Zhukov (2007),Solar and interplanetary sources of major geomagnetic storms (Dst < −
100 nT) during1996-2005,
J. Geophys. Res. Space Physics , , A10102, doi:10.1029/2007JA012321.Zhang, Y., W. Sun, X. S. Feng, C. S. Deehr, C. D. Fry, and M. Dryer (2008), Statisticalanalysis of corotating interaction regions and their geoeffectiveness during solar cycle 23, J. Geophys. Res. Space Physics , , A08106, doi:10.1029/2008JA013095.Zhao, L., T. H. Zurbuchen, and L. A. Fisk (2009), Global distribution of the solar windduring solar cycle 23: ACE observations, Geophys. Res. Lett. , , L14104, doi:10.1029/2009GL039181., L14104, doi:10.1029/2009GL039181.