In fl uence of coronal mass ejections on parameters of high-speed solar wind : a case study

This is anOp Abstract – We investigate the case of disagreement between predicted and observed in-situ parameters of the recurrent high-speed solar wind streams (HSSs) existing for Carrington rotation (CR) 2118 (December 2011) in comparison with CRs 2117 and 2119. The HSSs originated at the Sun from a recurrent polar coronal hole (CH) expanding to mid-latitudes, and its area in the central part of the solar disk increased with the rotation number. This part of the CH was responsible for the equatorial flank of the HSS directed to the Earth. The time and speed of arrival for this part of the HSS to the Earth were predicted by the hierarchical empirical model based on EUV-imaging and the Wang-Sheeley-Arge ENLIL semi-empirical model. The predicted parameters were compared with those measured in-situ. It was found, that for CR 2117 and CR 2119, the predicted HSS speed values agreed with the measured ones within the typical accuracy of ±100 km s . During CR 2118, the measured speed was on 217 km s 1 less than the value predicted in accordance with the increased area of the CH.We suppose that at CR 2118, the HSS overtook and interacted with complex ejecta formed from three merged coronal mass ejections (CMEs) with a mean speed about 400 km s . According to simulations of the Drag-based model, this complex ejecta might be created by several CMEs starting from the Sun in the period between 25 and 27 December 2011 and arriving to the Earth simultaneously with the HSS. Due to its higher density and magnetic field strength, the complex ejecta became an obstacle for the equatorial flank of the HSS and slowed it down. During CR 2117 and CR 2119, the CMEs appeared before the arrival of the HSSs, so the CMEs did not influence on the HSSs kinematics.


Introduction
Among various solar wind disturbances, high speed streams (HSSs) of solar wind originating from coronal holes (CHs) represent one of the main drivers of geomagnetic storms (Tsurutani et al., 1995(Tsurutani et al., , 2006;;Yermolaev et al., 2005Yermolaev et al., , 2009;;Echer et al., 2011a,b;Myagkova et al., 2013).Numerous studies have shown that geoeffectiveness of HSSs is determined by their speed, duration and magnetic field parameters.It is well established that there is a direct connection between the passage of the low and mid-latitude parts of CHs across the solar disk and the recurrent high-speed streams recorded at the Earth orbit (Nolte et al., 1976;Harvey & Sheeley, 1978;Wang & Sheeley, 1990).
During propagation in the solar corona and heliosphere, HSSs can interact with the ambient slow solar wind and other streams like coronal mass ejections (CMEs).Interaction of recurrent HSSs with slow solar wind in the heliosphere can generate corotating interaction regions (CIRs), which often produce moderate geomagnetic storms (Alves et al., 2006).Interaction of HSSs with CMEs is of particular interest for the space weather forecasting because it can strengthen or weaken geoeffectiveness of the event.Influence of different coronal structures as streamers, CHs, other CMEs, and the ambient solar wind on propagation of CMEs are the subject of many studies (see, for instance, papers by Gopalswamy et al., 2004;Mishra et al., 2015;Möstl et al., 2015 and references therein).Authors of several works (Gopalswamy et al., 2009;Mohamed et al., 2012;Nieves-Chinchilla et al., 2013) showed that presence of CHs can change the CME trajectory.In particular, polar CHs can deflect CMEs to the equator.Behannon et al. (1991); Burlaga et al. (2002Burlaga et al. ( , 2003)); Rouillard et al. (2010) discussed existence of compound structures in the solar wind À merged interaction regions (MIRs) formed by the CME-HSS interaction and observed in-situ in the solar wind.The MIRs are relatively large structures with the radial extent ≈2/3 AU and 3/4 AU and may include a complex ejecta, a magnetic cloud, shocks, a corotating stream and a heliospheric plasma sheet.Fenrich & Luhmann (1998) found that a fast solar wind stream could compress the preceding magnetic cloud, which results in increase of the plasma density and magnetic field strength providing a significant effect upon geomagnetic storm intensity if the magnetic cloud is of N-S polarity.However, the above mentioned papers did not consider influence of the CME-HSS interaction on the observed in-situ parameters of the HSS.
Currently several models are used for prediction of the solar wind parameters in the near-Earth space and beyond, which consider HSSs and CMEs as independent features.The quality of these models is evaluated by the disagreement between prediction and measurements of different solar wind parameters, among which the most important are the arrival time and speed at 1 AU (Zhao & Dryer, 2014;Jian et al., 2015;Reiss et al., 2016).The semi-empirical Wang-Sheeley-Arge model (WSA: Wang & Sheeley, 1990;Arge & Pizzo, 2000) based on the magnetic field expansion factor, the advanced Magnetohydrodynamic-based WSA-ENLIL prediction model (Odstrčil & Pizzo, 1999) and the empirical models based on the EUV-imaging (e.g. Vr snak et al., 2007;Krista & Gallagher, 2009;Shugay et al., 2011;Rotter et al., 2015) are most often used in operational practice.All types of the models use position and geometry of CHs for simulation of quasistationary streams of HSSs, but the EUV-imaging models are better following the short-time variations in the CH geometry, whereas the WSA and WSA-ENLIL models more effectively follow the global magnetic field configuration of the Sun.
CMEs originated at the Sun propagate through the heliosphere as expanding plasma structures reaching the Earth as interplanetary coronal mass ejections (ICMEs).Commonly two types of the models are often used for prediction of the arrival time and speed of ICMEs based on the near-Sun coronagraphic observations.Among the kinematic empirical models, the most advanced Drag-based model (DBM: Vr snak et al., 2013) consider that propagation of a CME in the heliosphere is governed by its interaction with the ambient solar wind depending on their relative speeds.A CME propagates with deceleration, if the stream velocity is higher than that of the ambient wind, or with acceleration in the opposite case.A modification of the WSA-ENLIL model, the WSA-ENLIL Cone model (Mays et al., 2015), provides simulation of a CME propagation in the heliosphere using information about initial parameters of CME from coronagraphic observations.This model describes expansion of the CME plasma in the heliosphere providing its speed and density as functions of time and distance in three dimensions, but likewise the kinematic models, it does not take into account probable magnetic interaction between HSS and CME and its influence on the solar wind parameters in the near-Earth space.Reiss et al. (2016) compared effectiveness of the empirical solar wind forecast (ESWF) model based on a relation between the areas of CHs as observed in EUV and the solar wind speed at 1 AU, and the WSA model.They found that 70% of the associated high-speed enhancements are correctly predicted by the ESWF model within ±100 km s À1 and 52% are correctly predicted within ±50 km s À1 for the period 2011-2014.In comparison, the WSA model correctly predicts 78% of the high-speed enhancements with the accuracy of ±100 km s À1 and 49% À within ± 50 km s À1 .The hierarchical empirical prediction model used in this work (Shugay et al., 2011, hereafter À the HE model) is similar to the ESWF tool in the principle of operation and has similar accuracy.
However, in some cases the disagreement between prediction and measurements surpasses the typical inaccuracy for mentioned models ∼100 km s À1 in two or more times.To improve the quality of the forecast, it is important to identify the causes of such strong disagreement.In particular, the HSS propagation anomalies can be caused by residual transient distortion of the background solar wind after passing a CME (Temmer et al., 2017).The authors investigated the impact of the detected in-situ ICMEs to the speed of the background solar wind predicted by different solar wind models (ESWF, WSA, persistence model) for the period 2011-2015.The observed solar wind speed within the ICME interval showed an increase of 18-32% above the expected background, and during the period of two days after the ICME displayed an increase of 9-24%.The total duration of enhanced deviations lasted over about three and up to six days after the ICME start, which is much longer than the average duration of an ICME disturbance itself.The studied solar wind disturbances surely should influence on the HSS propagation, although the authors did not mention it.One can suppose, that the largest disagreement between the predicted and measured arrival time and speed of a HSS may be caused by its direct interaction with a CME (or several CMEs) in the heliosphere.
In this work, we consider predictions of three HSSs occurred in the consecutive Carrington rotations (CRs) during November 2011-February 2012 (CRs 2117-2119), when the same recurrent polar CH expanded to middle latitudes crossed the central meridian.In our recent publication (Shugay et al., 2017) we mentioned this case as an example of strong disagreement between the HSS forecast and measurements, and here we give a detailed analysis of the events.According to the predictions by various models, the HSSs should be observed near the Earth in all rotations with the speed increasing due to the expanding middle-and low-latitude area of the CH.We consider disagreement between predictions and observations of the HSSs, which can be consequences of interaction between HSSs and cotemporal Earth-directed CMEs in the heliosphere.Our consideration is based on predictions by the HE model.We refer to simulations by the WSA-ENLIL Cone model only to demonstrate that the result is model independent.

Prediction of solar wind phenomena by the HE model
To predict arrival of HSS and transient CME streams to the Earth, one needs first to determine the sources of these streams, which can be identified by specific signatures.HSSs originate in CHs, which can be identified as large regions of open magnetic field lines or seen in the EUV wavelength range (typically, in the 19.3 nm line) as wide dark areas (Fig. 1, the upper panel).The procedure of the CH identification applied in the HE model has been described in Shugay et al. (2011Shugay et al. ( , 2014)); Slemzin & Shugai (2015).
To predict the arrival time and velocity of HSSs at 1 AU, we used as the initial data the EUV images in 19.3 nm band from the Advanced Imaging Assembly telescope on the Solar Dynamic Observatory (SDO/AIA) (Lemen et al., 2012) taken in the standard FITS-format, the sampling rate being one image per hour.In our model, we identify CHs by the intensity below the threshold value, which is calculated as I mean ⋅ k, where I mean is the average intensity of the solar disk without taking into account the limb brightening, and k is the empirical threshold coefficient, different for various wave length and spacecraft (Shugay et al., 2011).The regions of decreased intensity in the images of the Sun in the EUV wavelength range correspond not only to the regions of CHs, but also include the filament channels (FC).We excluded FCs by analysis of magnetic field distribution in the dimmed regions, which is different in CHs and FCs (Scholl & Habbal, 2008).To separate the CHs from FCs, we consider the data on the radial component of the photospheric magnetic field measured by the Helioseismic and Magnetic Imager on SDO (Hurlburt et al., 2012).
The solar wind sources positioned in the central part of the solar disk produce a measurable contribution to solar wind near the Earth.Therefore, we calculate the area of CHs within the source region confined by ±20°longitude from the central meridian of the Sun and by ±50°latitude from the solar equator.The HSS velocity is calculated as a linear function of the CH area (Slemzin & Shugai, 2015).We used this model for association of the passage of CH over the solar disk with the arrival of the HSS at 1 AU.The velocity of the solar wind as a function of time is calculated as follows: where S i ðt i 0 Þ is the CH relative area taken in the selected range in latitude and longitude at the time t i 0 at the wavelength l i , V min is the minimum (background) solar wind velocity, A i is a parameter depending on the wavelength l i , and t is the time of the HSS arrival to the near-Earth orbit according to the velocity prediction.The HE model calculates the HSS arrival time from the speed temporal profile with the transition time determined under the assumption that the stream speed in the heliosphere is constant (the ballistic approximation).The prediction of the HSS arrival time and speed at 1 AU using the above described model based on the SDO/AIA data is available in real time on the site of Skobeltsyn Institute of Nuclear Physics of Moscow State University (SINP/MSU)1 .
The accuracy of HE model was tested at the period January 2011-February 2012.The root-mean-square error (RMSE) of the forecast is given by the relation: where n À total number of observed and predicted pairs, V 0 À the observed HSS speed and V f À the forecasted speed.For the mentioned period, the HE model has RMSE s f = 105 km s À1 , and the correlation coefficient between the observed and predicted values equals to 0.45.It should be mentioned that these estimations were obtained from the difference between the predicted HSS speed and the real speed of the solar wind including CMEs.The HE model correctly predicted 73% of all solar wind streams with the speed more than 400 km s À1 with deviation from observations within the range ±100 km s À1 .For these simulations we used the speed of the background solar wind (V min ) equal to 300 km s À1 , as the most probable value of the minimum speed of the solar wind during the period 2011-2012.CMEs originate due to spontaneous solar activity phenomena associated with X-ray flares and flux rope ejections in active regions (ARs), or due to gradual filament/prominence eruptions without flares outside ARs.Such signatures as fast loop expansion, dimmings, coronal waves or post-eruption arcades can localize their solar sources.To link ICMEs with CMEs directed to the Earth, we used the data from the COR2 coronagraphs on two Solar Terrestrial Relations Observatories (STEREO-A and B, Howard et al., 2008), presented in the Sun-Earth Connection Coronal and Heliospheric Investigation instrument database2 and C2 instrument of Large Angle and Spectrotrometric Coronagraph (LASCO: Brueckner et al., 1995) onboard the Solar and Heliospheric Observatory (SOHO: Domingo et al., 1995), presented in the Coordinated Data Analysis Workshops (CDAW) CME list3 .To study solar origins of the identified CMEs we used EUV images from the SDO/AIA telescope4 and the Solar Demon database5 of the event detection by flares, dimmings and EUV waves.For calculations of the CME arrival time we used two heliospheric CME-propagation models: DBM6 (Vr snak et al., 2013) and the WSA-ENLIL Cone model (Odstrčil & Pizzo, 1999;Pizzo et al., 2011) 7 , both available online.Below, we will denote the data obtained from STEREO-A as the STA data and those from STEREO-B as STB data.In the period under study STEREO-A and B were located near the ecliptic plane at the angles 106-108°and (À105) to (À115)°relative to the Sun-Earth line.

Identification of the solar wind phenomena in-situ
Solar wind phenomena are typically identified using the standard criteria for the solar wind parameters (Sheeley & Harvey, 1981;Burlaga, 1991;Lepping et al., 2003;Zurbuchen & Richardson, 2006;Yermolaev et al., 2009;Borovsky & Denton, 2010;Mitsakou & Moussas, 2014;Prise et al., 2015).The solar wind features observed in connection with the HSS propagation are CIRs and HSSs (body): CIR is an interaction region between the slow and fast solar wind streams and characterized by an increase and then a decrease in the proton density, an increase in the velocity and kinetic temperature of the protons, an increase in the strength of the interplanetary magnetic field; HSS has the bulk speed V > 450 km s À1 (in the rise of Cycle 24 V > 400 km s À1 , Rod 'kin et al., 2016) and low values of the proton density and magnetic field.Propagation of a CME can cause the signatures of such phenomena as interplanetary shock (IS), sheath (SH), and ICME.Interplanetary shock is a sudden increase of the solar wind speed on more than 20 km s À1 accompanied with jumps in the plasma and magnetic field parameters; SH is a compression region before fast ICME an enhanced proton temperature, density and beta parameter.Numerical criteria and procedure of identification are described in details by Yermolaev et al. (2009), the catalog of the identified solar wind phenomena for 1976-2016 is presented at the web-site of Space Research Institute, Moscow (IKI solar wind database) 8 .
By analysis of the data of the OMNI dataset9 for the period 1976-2000 and using the method of the double superposed epoch analysis for large numbers of events (Yermolaev et al., 2010), we classified and parameterized temporal sequences of phenomena in assumption of absence of interaction with other streams (Yermolaev et al., 2015).As a result, the averaged temporal profiles for the main parameters of solar wind and IMF have been obtained.Similar methods of the solar wind profile analysis were effectively used in the papers by Yokoyama & Kamide (1997); Lepping et al. (2003); Kilpua et al. (2015).It should be noted that this method does not take into account the situations when different solar wind phenomena follow one after another and interact in the heliosphere.Thus, difference between measured and standard parameters of the classified phenomena can indicate such interaction in each particular case.
We analyzed solar wind parameters using the data from the Advanced Composition Explorer (ACE) and the WIND spacecraft for the period November 2011-February 2012.The hourly averaged values of the IMF and the solar wind parameters were taken from the instruments Magnetic Field Experiment (MAG: Smith et al., 1998) and Solar Wind Electron, Proton, and Alpha Monitor (SWEPAM: McComas et al., 1998) on ACE and the Solar Wind Experiment on WIND (SWE: Gloeckler et al., 1995).We used the ACE hourly averaged data (proton speed, magnetic field, He/H ratio) supplemented by the WIND hourly data (proton density, proton temperature) in the ACE data gaps.The beta parameter was calculated from the measured proton density, temperature, and magnetic field strength.

Results and discussion
In the next sections, we consistently describe observations of Sun, measurements of the solar wind parameters at 1 AU and their comparison with the model predictions in the period from the end of November 2011 to the beginning of February 2012 (CRs 2117-2119).

Prediction of HSSs and comparison with in-situ observations
During CRs 2117-2119, the potential solar sources of HSS were the large recurrent southern polar CH extended to the middle latitudes and, with smaller contribution, the growing high-latitude northern polar CH. Figure 1 (upper panel) shows the locations of the CHs in the SDO/AIA 19.3 nm solar images at CRs 2117-2119.The bottom panel shows the temporal variation of the CH areas in the source region within the ±20°l ongitude and ±50°latitude.Parameters of the HSS solar sources, predicted by the HE model and observed solar wind speed and the arrival times in UT are listed in Table 1. Figure 2 shows the temporal dependence of the predicted solar wind speed calculated by the HE model and by the WSA-ENLIL Cone model in comparison with the speed measured by ACE.The mid-latitude CHs produce the solar wind with the moderate speed (typically, 450-550 km s À1 , in Cycle 24-above 400 km s À1 ), low density and the IMF strength in the body of the stream after passage of CIRs.
Figure 1 shows that the southern CH at CR 2117 extended to the solar equator to ∼30°and at CR 2118 to 25°of the southern latitude.At CR 2117, the CH area have two maxima due to two parts of the CH area that expand below 50°of the southern latitude (Fig. 1), so it produced two HSSs À HSS 1 and HSS 2 (Tab. 1) with the predicted speeds corresponding to the value of the maximum CH area.In accordance with the observed geometry of the CH during all three CRs, the HE model predicted occurrence of the HSS near the Earth with the speed growing with the rotation number (Tab.1).
Figure 2 shows the ACE measurements of the solar wind speed for CRs 2117-2119 (the black line) in comparison with predictions by the HE model (the yellow line) and the WSA-ENLIL Cone model (the green line).In the period from 29 November 2011 to 7 December 2011 (CR 2117), the solar wind speed (black line) has two broad maxima: from 30 November 2011, 18:00 UT to 2 December 2011, 10:00 UT (HSS 1), and from 3 December 2011, 13:00 UT to 7 December 2011, 00:00 UT (HSS 2) associated with two parts of mid-latitude extensions of the southern CH.According to the data in Table 1, the observed maximal values of the solar wind speed agree with predictions by the HE model within the uncertainty less than 60 km s À1 .The WSA-ENLIL Cone model showed larger disagreement within ±100 km s À1 in the period 2-3 December 2011, probably, associated with the erroneous prediction of a CME arrival in the tail of HSS 1.
In the period from 28 December 2011, 06:00 UT to 7 January 2012, 19:00 UT (CR 2118), the HE model predicted the HSS arrival on 30 December 2011 at 06:00 UT with the speed of 597 km s À1 , but really the solar wind with some signatures of HSS (see in the next Section) was observed starting from 31 December 2011, 00:00 UT, with the speed 380 km s À1 , that is on 217 km s À1 less than it was predicted.Such disagreement between the predicted and measured arrival times and speeds of the HSS sufficiently exceeds the typical uncertainty of the HE model.A decrease of the HSS speed at CR 2118 in comparison with that at CR 2117 cannot be explained by variation of the CH geometry, because at CR 2118 the CH area in the source region increased, and its upper boundary shifted towards the equator.
At CR 2119, according to Figure 1, the boundary of the recurrent CH was located at 20°southward from the equator, and its geoeffective area increased in comparison with the preceding rotations due to the part of the northern CH.The HE model predicted appearance of the HSS in the period from 27 January 2012, 05:00 UT to 31 January 2012, 12:00 UT with the maximal speed of 634 km s À1 .The observed speed of the HSS has reached 550 km s À1 , which is on 84 km s À1 less than the predicted value and lies in the accuracy limits of the HE model.
We suggest that the significant disagreement between prediction and observation of the HSS at CR 2118 might be caused by interaction of the HSS with the CMEs propagated in the heliosphere towards the Earth at the same period.In the next sections, we investigate in detail the solar wind parameters around the periods of the HSS prediction and the solar phenomena, which can influence on the HSS propagation.

Solar phenomena and their influence on the HSS parameters
Table 2 presents information about CMEs started during the period, when the recurrent CH was present in the source region at the solar disk, and predictions on their arrival to the Earth.These CMEs were selected from the STEREO and LASCO CME databases as it was described in the Section 2. Their arrival time at 1 AU and speed were estimated using DBM (the speed of the background solar wind was taken to be 300 km s À1 , similar to that taken for the HSS simulations with the HE model).The solar wind measurements (Fig. 3) show a shock on 28 November 2011, 21:00 UT followed by a SH region of enhanced density, temperature and magnetic field (28 November 2011, 22:00 UT-29 November 2011, 18:00 UT), an ICME from 29 November 2011, 19:00 UT to 30 November 2011, 17:00 UT and the HSS signatures from 30 November 2011, 18:00 UT to 2 December 2011, 10:00 UT.This complex structure represents a MIR, when two or several CMEs merge with HSS (Burlaga et al., 2003;Rouillard et al., 2010 and references therein).In our case, MIR lasting from 28 November 2011, 22:00 UT to 30 November 2011, 17:00 UT included a shock, a SH, an ICME and CIR followed by the HSS body.The snapshot of the WSA-ENLIL Cone modeling (Fig. 4) shows that CME 1 overtook HSS 1 in the heliosphere and arrived to the Earth practically at the same time and with the same speed.
After HSS 1, the solar wind data contains signatures of the second ICME from 2 December 2011, 11:00 UT with the speed 390 km s À1 .Most probably, this ICME corresponds to CME 2 started from the Sun on 28 November 2011, 11:24(STA) UTwith the speed 322 km s À1 .According to the DBM simulation, CME 2 was predicted to arrive on 2 December 2011, 19:00 UT, with a speed 320 km s À1 that corresponds to the speed of the observed ICME within the DBM accuracy.The CH corresponding to the HSS 2 reached its maximum size two days later, 30 November 2011, 06:00 UT, and arrived on 3 December 2011, 13:00 UTwith the speed 469 km s À1 .Thus, these streams propagating one after another did not interacted in the heliosphere, so they did not generated a shock and a MIR.
At CR 2118, according to the predictions by the HE model (Tab.1), the CH reached its maximum area on 26 December 2011 at 11:00 UT, and the corresponding HSS should arrive to the Earth on 30 December 2011, 06:00 UT with the speed of 597 km s À1 .The ACE data in Figure 5 show that during the period from 28 December 2011, 03:00 UT and 30 December 2011, 06:00 UT, there was a period of strong disturbances in the solar wind, marked in Figure 5 by the red horizontal bar, which included large enhancements in the proton density, speed, temperature and magnetic field strength.After these disturbances, until 2 January 2012, 04:00 UT, there was a short ICME, marked by the green bar, followed by the short-time enhancement in density, temperature and magnetic field strength, which may be identified as a stream interface region.From 31 December 2011, 00:00 UT to 2 January 2011, 04:00 UT there is a period of the turbulent solar wind, in which such plasma parameters, as proton temperature, magnetic field and beta, intermittently displayed signatures of HSS and ICME.In Figure 5, this period of the solar wind is marked by the blue dashed bar.As it was noted in Section 3.1, during this period the maximum speed of the HSS did not exceed 380 km s À1 , which is 217 km s À1 less than the predicted one.
Such feature, which includes the complex ejecta, the ICME, CIR and MIR like that identified at CR 2117, but with the more pronounced signatures.In particular, the proton density reached 47 cm À3 versus 20 cm À3 in the previous case, the proton temperature À 1 Â 10 5 K versus 5Â10 5 K. Observation of such feature suggests that the HSS met and interacted in the heliosphere with one or several CMEs.
By analysis of the probable CMEs observed by STA, STB and LASCO in the period from 25 to 27 December 2011, we identified, as the probable solar sources, several CMEs starting in the period between 25 and 27 December.Three directed to the Earth brightest CMEs observed by STA, STB and LASCO are listed in Table 2.These CMEs were also traced in the heliosphere by the High resolution Imagers on STA and STB 10 .DBM predicted that these CMEs taken separately might arrive to the Earth in the period from 28 December 2011, 13:00 UT to 30 December 2011, 06:00 UT with the speeds 390-420 km s À1 .Figure 6 shows a snapshot of the WSA-ENLIL Cone simulation of the solar wind conditions at the Earth on 29 December 2011, 00:00 UT, in the ecliptic and meridional planes.It occurs that these CMEs merged, created the complex ejecta and interacted with the faster HSS, which overtook them in the heliosphere.The WSA-ENLIL Cone simulation also shows that the main flow of the HSS propagated southward from the ecliptic plane, and only its small northern flank (in the meridional plane) moved in the direction to the Earth.We suppose that due to higher density and stronger magnetic field of the complex ejecta, its interaction with the HSS flank resulted in creation of a stratified structure with mixed HSS and CME plasma moving with the speed of the densest component À the slow CME complex.Probably, this effect might be caused by the physical mechanism as in the case of interaction of two CMEs described by Temmer et al. (2012), in which the fastest CME was decelerated by the slowest one.
Agreement of the measured HSS arrival time with the predicted one suggests that this interaction occurred close to the Earth.After 2 January 2012, 04:00 UT, the HSS signatures in the solar wind changed for the signatures of the second  2. At CR 2119 (Fig. 7), the HSS was predicted during the period from 27 January 2012, 05:00 UT to 31 January 2012, 12:00 UT with the speed 565 km s À1 .A day earlier, on 23 January 2012 about 04:00 UT a fast CME started with the speed exceeding 1000 km s À1 , which was predicted by DBM to arrive on 25 January at 17:00 UT with the speed of 480 km s À1 .In contrary with the previous rotations, at CR 2119 the HSS-CME interaction was not seen in the solar wind, so the arrival time and maximum speed of HSS predicted by the HE model agreed with the measurements.In the end of the HSS period, on 30 January 2012, 16:00 UT, WIND observed a fast forward shock in the solar wind (see the Heliospheric Shock Database11 ).Analysis of the CME events at the Sun in the end of January 2012 has shown that STA and STB did not registered any CMEs directed to the Earth, which could produce this shock.However, on 27 January 2012 at 18:27 UT, LASCO registered near the western limb a halo CME with the speed 2500 km s À1 , which was associated with the X-ray flare X 1.7 observed by GOES.Although this powerful CME propagated towards STA in the direction perpendicular to the Sun-Earth line, it produced the global shock detected by WIND without the ICME signatures afterwards.

Summary
We present a case study of significant disagreement between the predicted and observed parameters of HSSs of solar wind revealed in CR 2118 -the second of three consecutive appearances of the recurrent CH in November 2011-February 2012 (CRs 2117-2119), when propagation of the HSSs in the heliosphere were accompanied by CMEs on the Sun.For prediction, we used two models.The first one is the kinematical hierarchical empirical model (HE model: Shugay et al., 2011Shugay et al., , 2014) ) based on a relation between the HSS speed and area of the CH in the central part of the solar disk measured from the SDO/AIA images in the 19.3 nm band.The second is the WSA-ENLIL Cone model (Mays et al., 2015), which uses as the initial data for the HSS prediction the magnetic field expansion factor dependent on the position and area of the open magnetic field, and for the CME prediction À the coronagraph data.Both models predict an increase of the HSS speed with the CR number due to the growing CH area in the central part of the solar disk, not taking into account probable interaction between the HSSs and the CMEs simultaneously propagated in the heliosphere.
From analysis of the solar wind parameters measured insitu by ACE and WIND, we found that at CR 2117 and CR      3, for the interval 25 January 2012, 00:00 UT-05 February 2012, 12:00 UT (CR 2119).with the difference of 57 km s À1 and 82 km s À1 .At CR 2118, instead, the measured HSS speed was on 217 km s À1 less than the predicted value.We suppose that the low speed of the solar wind in the place of predicted HSS at CR 2118 might be caused by interaction with the complex of CMEs propagated at the same period in the heliosphere.By analysis of the events presented in the STEREO and LASCO CME lists, we selected the CMEs propagated towards the Earth, which might interact with the HSSs during their passage in the heliosphere.
We found that at CR 2117, nine hours after the slow HSS (V ∼ 425 km s À1 , a fast CME started with the speed of ∼1000 km s À1 , overtook the HSS near the Earth, and the both streams arrived to the Earth simultaneously with similar speeds of about 450 km s À1 .Because of this interaction, the HSS signatures in the solar wind appeared later than the predicted arrival time, but due to similarity of the stream speeds, the HSS speed was very close to the predicted value.
At CR 2118, instead, the faster HSS overtook the slower moving complex ejecta originated from the series of merged CMEs, which arrived to the Earth just before the HSS.We suggest that due to its higher density and magnetic field strength, the complex ejecta represented an obstacle for the HSS and slowed down its equatorial flank on 217 km s À1 .
At CR 2119, the closest in time CME directed to the Earth arrived ahead of the HSS.In contrary with the previous rotations, at CR 2119, the signatures of the HSS-CME interaction were absent, so the HSS speed predicted by the HE model agrees with the solar wind measurements within the typical error of ±100 km s À1 .
In conclusion, our study shows the case, when the HSS-CME interaction in the heliosphere affected on the HSS propagation, which resulted in disagreement between the HSS parameters measured in-situ and predicted by the models based on the CH geometry.The effect of such interaction depends on the 3D geometry and propagation directions of the HSS and CMEs as well as on the relations between their timing and speeds.Probably, this interaction has similar physical nature as in the case of deceleration of the fast CME by the slower one described by Temmer et al. (2012).We suggest that significant disagreement between predictions and measurements of HSSs in-situ at 1 AU exceeding the accuracy of the used model may serve as an indicator of strong HSS-CME interaction for further investigation.

Fig. 1 .
Fig. 1.Top: images of the Sun at wavelength 19.3 nm from SDO/AIA for CR 2117, 2118 and 2119 (from left to right on 27 November 2011, 28 December 2011 and 25 January 2012).Bottom: temporal variation of the CH area (in relative units) for three CRs.The abscissa is relative time in hours.
Figures 3, 5 and 7 demonstrate the solar wind parameters measured in-situ near the periods of the predicted HSS arrivals.At CR 2117, according to two maxima in the temporal dependence of the CH area (see Fig. 1), two HSSs-HSS 1 and HSS 2 were formed at the Sun.HSS 1 reached its maximal speed on 25 November 2011, 23:00 UT.The HE model predicted its arrival to the Earth in the period from 29 November 2011, 12:00 UT to 3 December 2011, 10:00 UT with the speed 425 km s À1 .Nine hours after the CH areas reached its maximum size, a halo CME 1 started on 26 November 2011, at 08:24(STA)/07:55(STB) UT with the speed 1139/821 km s À1 .DBM (Tab.2) predicted the arrival of CME 1 on 29 November 2011, 02:00 UT with the speed of 460 km s À1 .

Fig. 2 .
Fig. 2. Data presented from top to bottom for CRs 2117-2119.The proton speed measured by ACE (the black curve), the proton speed calculated by the HE model (the yellow curve) and calculated by the WSA-ENLIL Cone model (the green curve).

Fig. 3 .
Fig.3.Solar wind parameters for the period 27 November 2011, 00:00 UT-08 December 2012, 12:00 UT (CR 2117).The dash-dotted lines implicate the predicted period of the HSS arrival to the Earth.The dotted lines correspond to: in the proton temperature graph À the expected temperature, in the magnetic field graph À B z component.The solid vertical line shows a shock.The solar wind phenomena marked by the color bars are: HSSs (blue bar), ICMEs (green bar), CME-HSS interaction regions (red bar).

Fig. 6 .
Fig. 6.Snapshot of the WSA-ENLIL Cone simulation of the solar wind conditions in the heliosphere on 29 December 2011, 00:00 UT in the ecliptic plane (a) and meridional plane (b).

Table 1 .
Times of maximum CH area, predicted and observed parameters of HSSs.