Testing the background solar wind modelled by EUHFORIA
J. Hinterreiter, J. Magdalenic, M. Temmer, C. Verbeke, I.C. Jeberaj, E. Samara, E. Asvestari, S. Poedts, J. Pomoell, E. Kilpua, L. Rodriguez, C. Scolini, A. Isavnin
SSolar PhysicsDOI: 10.1007/ ••••• - ••• - ••• - •••• - • Assessing the performance of EUHFORIA modelingthe background solar wind
J¨urgen Hinterreiter · Jasmina Magdalenic · Manuela Temmer · Christine Verbeke · ImmanuelChristopher Jebaraj · Evangelia Samara · Eleanna Asvestari · Stefaan Poedts · Jens Pomoell · Emilia Kilpua · Luciano Rodriguez · Camilla Scolini · Alexey Isavnin (cid:66) J. Hinterreiter [email protected]
J. Magdalenic [email protected]
M. Temmer [email protected]
C. Verbeke [email protected]
I. C. Jebaraj [email protected]
E. Samara [email protected]
E. Asvestari eleanna.asvestari@helsinki.fi
S. Poedts [email protected]
J. Pomoell jens.pomoell@helsinki.fi
E. K. J. Kilpua emilia.kilpua@helsinki.fi
L. Rodriguez [email protected]
C. Scolini [email protected]
A. Isavnin [email protected] Space Research Institute, Austrian Academy of Sciences, Graz, Schmiedlstraße 6,8042 Graz, Austria Institute of Physics, University of Graz, Universit¨atsplatz 5, 8010 Graz, Austria
SOLA: Paper_Euhforia.tex; 19 August 2019; 2:43; p. 1 a r X i v : . [ a s t r o - ph . S R ] A ug interreiter et al. © Springer ••••
Abstract
In order to address the growing need for more accurate space weatherpredictions, a new model named EUHFORIA (EUropean Heliospheric FORe-casting Information Asset) was recently developed (Pomoell and Poedts, 2018).We present first results of the performance assessment for the solar wind model-ing with EUHFORIA and identify possible limitations of its present setup. Usingthe basic EUHFORIA 1.0.4. model setup with the default input parameters, wemodeled background solar wind (no coronal mass ejections) and compared theobtained results with ACE, in situ measurements. For the need of statisticalstudy we developed a technique of combining daily EUHFORIA runs into con-tinuous time series. The combined time series were derived for the years 2008(low solar activity) and 2012 (high solar activity) from which in situ speed anddensity profiles were extracted. We find for the low activity phase a better matchbetween model results and observations compared to the considered high activitytime interval. The quality of the modeled solar wind parameters is found to berather variable. Therefore, to better understand the obtained results we alsoqualitatively inspected characteristics of coronal holes, sources of the studiedfast streams. We discuss how different characteristics of the coronal holes andinput parameters to EUHFORIA influence the modeled fast solar wind, andsuggest possibilities for the improvements of the model.
Keywords:
Coronal Holes; Magnetic fields, Models; Solar Wind; Magnetohy-drodynamics
1. Introduction
The solar wind is a continuous flow of charged particles propagating radiallyoutward from the hot corona of the Sun into interplanetary space. The speedmeasured at 1 AU heliocentric distance covers generally a range between 300and 800 km s − , consisting of slow solar wind and of high speed solar windstreams that have different characteristics and sources (e.g., Cranmer, Gibson,and Riley, 2017; Schwenn, 2006).The sources of the slow solar wind are closed magnetic field regions of coronalloops, active regions, coronal hole (CH) boundaries, but also streamers andpseudostreamers (Cranmer, Gibson, and Riley, 2017). On the other hand, fastsolar wind emanates from open magnetic field regions, CHs, along which ionizedatoms (mainly protons and alpha-particles) and electrons may easily escape to Solar–Terrestrial Centre of Excellence—SIDC, Royal Observatory of Belgium, 1180Brussels, Belgium Centre for mathematical Plasma Astrophysics (CmPA), KU Leuven, 3001 Leuven,Belgium Department of Physics, University of Helsinki, P.O. Box 64, 00014 Helsinki, Finland
SOLA: Paper_Euhforia.tex; 19 August 2019; 2:43; p. 2
UHFORIA background solar wind modeling interplanetary space. CHs are localized regions of low density and low temper-ature in the solar corona that are generally slowly evolving and may persistfor several solar rotations (Schwenn, 2006). However, where exactly within theCH the high speed component of the solar wind gets accelerated is not wellunderstood and is subject of numerous studies.High speed streams from CHs interact with the slower solar wind aheadcausing compression regions that can lead to geomagnetic storms and the faststream following the compression region with Alfvenic fluctuations can prolongsubstantially the recovery phase of the storm (e.g., Tsurutani and Gonzalez,1987). It is well acknowledged that during the maximum phase of the solar cyclespace weather is affected mostly by transient coronal mass ejections (CME; e.g.,Webb and Howard, 2012), however, during the declining and minimum activ-ity phases high speed streams have significant impact (Tsurutani et al. , 2006;Richardson and Cane, 2012; Kilpua et al. , 2017). At all phases of solar cycle, highspeed solar wind streams have also a paramount impact causing enhancementsof Van Allen belt electron fluxes to relativistic electrons (e.g., Paulikas andBlake, 1979; Jaynes et al. , 2015; Kilpua et al. , 2015) and they strongly structureinterplanetary space which is an important factor when studying and forecastingthe propagation of CMEs. In general the morphology, area and location of CHsplay a major role in the properties of the resulting compression region, durationand speed of the fast stream, and thus, its space weather impact level (e.g.,Vrˇsnak, Temmer, and Veronig, 2007; Garton, Murray, and Gallagher, 2018). Forexample, statistical studies have shown that the equatorial parts of CHs are themain contributors to the fast solar wind streams measured at Earth (see e.g.,Karachik and Pevtsov, 2011; Hofmeister et al. , 2018) and that the speed of thesolar wind at Earth increases with increasing CH area (e.g., Rotter et al. , 2012;Nakagawa, Nozawa, and Shinbori, 2019). We note that with the evolution of aCH over time, also associated, the in-situ measured solar wind parameters canchange (e.g., Heinemann et al. , 2018).Models simulating the background solar wind are based on various methods,e.g., physics-based algorithms such as ENLIL (Odstrˇcil and Pizzo, 1999) or MAS(Linker et al. , 1999) using synoptic photospheric magnetic field maps as input,empirical relations between observed areas of CHs and measured solar windspeeds at 1 AU (Vrˇsnak, Temmer, and Veronig, 2007; Rotter et al. , 2012), orsimple persistence models using in-situ measurements shifted forward by vari-able time-spans depending on the spacecraft location (e.g. Opitz et al. , 2009;Owens et al. , 2013). The performances of all the different solar wind models incomparison to actual measurements, reveal on average root-mean-square-errorsof around 100–150 km s − in the wind speed and time shifts in the arrival of thepeak speed of about ± ± et al. , 2008;MacNeice, 2009; Gressl et al. , 2014; Reiss et al. , 2016; Jian et al. , 2015; Temmer,Hinterreiter, and Reiss, 2018). In general, model performances decrease withincreased solar activity phases as CMEs frequently disturb the interplanetaryspace. Especially empirical solar wind models are not able to cope with thosedisturbances, but also for numerical models preconditioning is an importantaspect which needs to be taken into account (Temmer et al. , 2017). SOLA: Paper_Euhforia.tex; 19 August 2019; 2:43; p. 3 interreiter et al.
In order to address the growing need for more accurate space weather predic-tions, a new model named EUHFORIA (EUropean Heliospheric FORecastingInformation Asset) was recently developed (Pomoell and Poedts, 2018). In thefollowing we present the first performance assessment of the solar wind modeland identify possible caveats related to complex solar surface situations.
2. Solar wind modeling with EUHFORIA
EUHFORIA is a physics-based simulation tool consisting of three essential parts:a coronal model, a heliospheric model and an eruption model. The main pur-pose of the coronal model is to provide realistic plasma conditions of the solarwind at the interface radius r = 0.1 AU between the coronal and heliosphericmodel. The heliospheric model computes the time-dependent evolution of theplasma from the interface radius by numerically solving the MHD equationswith the boundary conditions provided by the coronal model. For simulatingtransient events, CMEs are injected at the interface radius of the eruptionmodel. Presently EUHFORIA shares similarities to the well-established solarwind/ICME model for the inner heliosphere, i.e. WSA-ENLIL (Odstrcil, Riley,and Zhao, 2004). An important feature of EUHFORIA is its flexibility. Thethree models, heliospheric, coronal and eruption one are fully autonomous andeach part of EUHFORIA can be easily substituted with other models (moredetails can be found in Pomoell and Poedts, 2018; Scolini et al. , 2018). Incontrast to ENLIL which gives the background solar wind parameters for afull Carrington rotation, EUHFORIA provides daily runs from hourly updatedstandard synoptic GONG magnetograms. In this way the central part of themagnetogram, used by EUHFORIA, is daily updated. For the purpose of thestatistical studies and easier comparison with in situ observations we combinedaily runs in order to obtain single time series (for a detailed description seeSection 2.2).In the present study we used EUHFORIA 1.0.4 version of the model, andwe focus on the coronal and heliospheric model, in order to assess how wellEUHFORIA simulates the background solar wind. For this study, we consideredtwo phases of solar activity, one year during minimum in 2008 and another yearduring maximum in 2012. As this is the first study of the solar wind modeling with EUHFORIA, weemployed the so-called default setup that uses default values for the inputparameters. For the coronal part of the model, we use synoptic magnetogramsfrom the Global Oscillation Network Group (GONG), and the potential fieldsource surface (PFSS) model (Altschuler and Newkirk, 1969) to simulate themagnetic field up to heights of 2.6 R (cid:12) (so called source surface height). This is We refer to ENLIL runs that can be performed at the Community Coordinated ModelingCenter (CCMC) which can be found under https://ccmc.gsfc.nasa.gov
SOLA: Paper_Euhforia.tex; 19 August 2019; 2:43; p. 4
UHFORIA background solar wind modeling
Figure 1.
Snapshot of the background solar wind radial speed modeled by EUHFORIA. Thetop left panel shows the MHD solution in the heliographic equatorial plane, and the right panelshows the meridional plane cut that includes the Earth (blue circle). The lower panel showscomparison of the modeled and observed solar wind by EUHFORIA and ACE, respectively. combined with the Schatten current sheet (SCS) model (Schatten, Wilcox, andNess, 1969) starting from the height of 2.3 R (cid:12) and that extends up to 0.1 AU.By overlapping the two models, a smoother transition between the lower coronalPFSS and upper coronal SCS model is obtained (see Pomoell and Poedts, 2018;McGregor et al. , 2008). To determine the solar wind plasma parameters at theinner boundary of the heliospheric model we use the empirical Wang-Sheely-Argemodel (Arge et al. , 2003) which is described below.In EUHFORIA the solar wind speed depends on several parameters and thefunctional form of the empirical relation can be selected by the user. In thisstudy we have employed the expression in the form: v ( f, d ) = v + v (1 + f ) α (cid:2) − . − ( d/w ) β (cid:3) , (1)where f and d are the flux tube expansion factor and the great circular angulardistance from the footpoint of each open field line to the nearest CH boundary, re-spectively. The parameters in Eq. 1 are set to v = 240 km s − , v = 675 km s − , α = 0.222, β = 1.25 and w = 0.02 rad. For a more detailed description see Eq.2 in Pomoell and Poedts (2018). Since the original WSA relation is designed toprovide the wind speed at Earth, and as the solar wind continues to acceleratebeyond the inner boundary in the heliospheric MHD model, we have additionallysubtracted 50 km s − to avoid a systematic overestimate of the wind speed. Tocompensate for the solar rotation, which is not included in the magnetic field SOLA: Paper_Euhforia.tex; 19 August 2019; 2:43; p. 5 interreiter et al. model, we rotate the solar wind speed map at the inner boundary by 10 ° . Wehave also limited the minimum and the maximum solar wind speed at the innerboundary to 275 and 625 km s − , respectively (according to McGregor et al. ,2011). In addition to the wind speed, the remaining MHD variables need to bedetermined. While the topology of the magnetic field is directly obtained fromthe SCS model, the magnitude of the solar wind magnetic field is set to bedirectly proportional to the speed. The plasma number density is given by n = n fsw ( v fsw /v r ) , (2)with the number density of the fast solar wind n fsw = 300 cm − (see e.g.,Bougeret, King, and Schwenn, 1984; Venzmer and Bothmer, 2018), the fastsolar wind speed v fsw = 675 km s − and v r coming from the empirical speedprescription. The maximum value v fsw = 675 km s − is considered to be in thesolar wind plasma with a magnetic field of 300 nT. For more details see Eq. 4in Pomoell and Poedts (2018).Finally, we use a constant plasma thermal pressure of 3.3 nPa, at the innerboundary, that is in accordance with the fast solar wind temperature of about0.8 MK. The angular resolution of the daily runs in this study was 4 ° , while 512grid cells were chosen in the radial direction to cover the 0.1 to 2 AU domain.An example of the background solar wind speed modeled by EUHFORIA,for the time interval of seven days in March 2008, is presented in Figure 1.The two top panels (the heliographic equatorial and the meridional plane cutsplotted in the left and right panel, respectively) show that the Earth has entereda region of extended fast flow. The time of the snapshot is also marked by theblack vertical line in the bottom panel which shows a comparison between thein situ observations and modeled solar wind speed. For this time period, we notea good match between the modeled solar wind by EUHFORIA and the in situmeasurements (cf. bottom panel of Figure 1). For the systematic testing of the background solar wind, we used EUHFORIAdaily runs, i.e., model outputs with default parameters, based on standard syn-optic GONG magnetograms (the selected time was about 23:30 UT each day),for the complete years 2008 and 2012. We consider that each daily run, based onone magnetogram input, simulates the background solar wind at the heliocentricdistance of 1 AU over a total time span of 14 days ( ± ± ° inlongitude (see gray slice in Figure 2) with a temporal resolution of 10 minutes.The central region of the Sun has the magnetic field information with the lowestprojection effects, and is thus the most reliable part of the magnetogram. Tocombine the individual daily runs which overlap in time, we therefore developeda method containing information with highest weight on the central region ofthe Sun. The central region is defined as ± ° ) as given in the schematic drawing in Figure 2a. The weighting of each curveis done by a Gaussian distribution with the central part receiving the strongestweight (see Figure 2b). SOLA: Paper_Euhforia.tex; 19 August 2019; 2:43; p. 6
UHFORIA background solar wind modeling
Figure 2.
Schematic representation of combining EUHFORIA model output for consecutivedays. a) Different colors represent the selected range ( ± ° from the solar central region)for each day. Indicated in gray is the full range ( ± ° ) provided by the model. b) Gaussianweight used for the model properties shown for three individual days. In Figure 3 we demonstrate how the method was applied. The top panel ofFigure 3 shows the solar wind speed modeled by EUHFORIA for the full modeloutput ( ± ± ± ± SOLA: Paper_Euhforia.tex; 19 August 2019; 2:43; p. 7 interreiter et al.
Figure 3.
Solar wind speed from July to August 2008. Top panel: Full EUHFORIA modeloutput ( ± ± Figure 4.
Comparison of different shifts of the central region. The red curve (0d) representsthe central region used for the individual runs. −
3d indicates that the central region is shifted3 days to the East while +3d indicates a shifting of the central region to the West.
SOLA: Paper_Euhforia.tex; 19 August 2019; 2:43; p. 8
UHFORIA background solar wind modeling
3. Comparison of in-situ observations and modeled solar wind
Figure 5.
EUHFORIA model output (red: velocity, blue: density) in comparison to in-situmeasurements (gray) for 2008. Top panel: Solar wind bulk velocity. Bottom panel: Solar winddensity.
Figure 6.
Same as Figure 5 but for the year 2012.
In order to asses the performance of the model we chose two intervals ofdifferent solar activity levels. At first, a quiet period during 2008 is considered,for which only three interplanetary coronal mass ejections (ICMEs), at the end ofthe year, were reported in the near-Earth solar wind according to the Richardsonand Cane ICME list (Richardson and Cane, 2010, see ). This period can serve as benchmarktime interval for the model performance as it almost optimally represents the
SOLA: Paper_Euhforia.tex; 19 August 2019; 2:43; p. 9 interreiter et al. background solar wind without significant transient perturbations. A secondconsidered interval covers the year 2012, a period with rather high level of solaractivity during which 35 ICMEs are reported (cf. Richardson and Cane ICMElist). In order to evaluate how well EUHFORIA models the background solarwind, we compare the combined time series (see Section 2.2) with the in-situmeasured plasma parameters speed and density as provided by the Solar WindElectron, Proton and Alpha Monitor onboard the Advanced Composite Explorer(SWEPAM/ACE, McComas et al. , 1998).Figures 5 and 6 show the results obtained for the years 2008 and 2012. Thegray curves represent observed values by ACE, while red and blue curves rep-resent modeled values of the solar wind speed and density, respectively. Thepresented statistics of the background solar wind modeled with EUHFORIAshows on average lower values of the modeled solar wind speed than the in-situ measured velocity. On the other hand the modeled solar wind density isconsiderably higher than the observed one. In the present setup of EUHFORIAthese two solar wind plasma parameters are coupled (cf. Eq. 2), and improvedmodeling of the solar wind speed will also result in a better modeled solar winddensity. We also noticed that the correlation between modeled and observedvalues is significantly better in the first half of year 2008 (Figure 5). In the secondhalf of year 2008, the maximum speeds for the fast solar wind speed are not wellmodeled by EUHFORIA, and also the minimum values are significantly different,i.e., larger than the observed ones. For the year 2012 the discrepancies betweenthe modeled values and observations are more pronounced. Nevertheless, periodsof lower wind speeds during 2012 are rather well reproduced, which might besimply a consequence of a very low wind speed in general obtained for this year.The in-situ solar wind speed, for both studied years, was also compared to theindividual daily runs in order to assess the probability of artificially enhancedor reduced fast wind flows due to combining of the daily runs (Section 2.3). Inthe two studied years we found only one case of the fast solar wind which wasobserved in the majority of the daily runs but not in the combined time series(around 22 August 2012). The opposite cases, where the combined time seriesshow significant increase of the solar wind speed that was not modeled in themajority of the relevant daily runs, were not found.As a consequence of the, on average, underestimated solar wind speed modeledby EUHFORIA, fast flows arrive with a systematic delay in time. The amount ofdelay depends on the difference between the modeled and observed wind speed.For example, the fast solar wind with average speed of 600 km s − will needabout 2.9 days to arrive to the Earth, while those of about 500 km s − will needabout 3.5 days. In this case the induced latency of modeled solar wind will beabout 14 hours. We observe the influence of this effect particularly strong in thesecond half of the year 2008 (Figure 5). In order to evaluate the EUHFORIA model performance we present a hit-missstatistics using two different methods for comparing measured and modeledresults. We also compare the minimum and maximum phase of the results and
SOLA: Paper_Euhforia.tex; 19 August 2019; 2:43; p. 10
UHFORIA background solar wind modeling give initial results on the effects of different input parameters for the model. Inthis analysis we focus only on the solar wind velocity.
To evaluate the model performance, we calculate continuous variables (e.g.,Root-mean-squared error RMSE) and apply an event-based approach for de-tecting the maxima (peak finding algorithm) in the solar wind observations. Forthe event-based approach we used an automatic peak finding algorithm. To bedefined as a peak, certain properties (minimum speed = 400 km s − , minimumgradient = 60 km s − , for further details see Reiss et al. , 2016) have to befulfilled. A hit is found, if the modeled peak appears within a time window of ± − . As can be seen from Figure 7, in 2008 (top panel), 39 solarwind peaks are detected in the EUHFORIA combined time series and 43 in thein-situ data. Applying the automatic peak finding algorithm method, we obtain18 hits, 21 false alarms and 25 misses. In 2012 (bottom panel in Figure 7), theEUHFORIA combined time series shows 21 peaks and 38 are detected in thein-situ observations. This corresponds to 14 hits, 7 false alarms and 24 misses.As this is a rather poor result we inspect the solar wind profiles (observed andmodeled) in more detail and investigate the reason of the poor performance. The in-situ observations frequently show several subsequent local maxima ofthe solar wind speed associated with a single fast flow generally originatingfrom a large and extended, in latitude or in longitude or both, CH. In sucha case the automatic peak finding algorithm finds several peaks and it is notpossible to make a one-to-one identification with the usually smooth increase ofthe solar wind speed modeled by EUHFORIA. In order to better understand suchlong lasting flows and to unambiguously relate modeled and observed velocitypeaks with each other, we checked the development of the CHs on the Suntwo days before and three days after the CH started its transition across thecentral meridian (see Figure 8). For this purpose we analyzed automatic CHareas detected by the CHIMERA software (Garton, Gallagher, and Murray,2018) and CH drawings (see Figure 8).As for the automatic method, the intervals corresponding to ICME arrivals,reported in a list by Richardson and Cane (2010) and observed in-situ, were
SOLA: Paper_Euhforia.tex; 19 August 2019; 2:43; p. 11 interreiter et al. J an - F eb - M a r- A p r- M a y - J un - J u l - A ug - S ep - O c t - N o v - D e c - D a t e [ m on t h - y ea r ]
300 400 500 600 700
Bulk speed [km/s] F o r e c a s t M ea s u r e m en t H i t ( ) F a l s e A l a r m ( ) M i ss ( ) I C M E a t A C E de t e c t ed A c e ( ) de t e c t ed E uh f o r i a ( ) M i n P k H e i gh t = J an - F eb - M a r- A p r- M a y - J un - J u l - A ug - S ep - O c t - N o v - D e c - D a t e [ m on t h - y ea r ]
300 400 500 600 700
Bulk speed [km/s] F o r e c a s t M ea s u r e m en t H i t ( ) F a l s e A l a r m ( ) M i ss ( ) I C M E a t A C E de t e c t ed A c e ( ) de t e c t ed E uh f o r i a ( ) M i n P k H e i gh t = F i g u r e . : E UH F O R I A m o d e l e d s o l a r w i ndbu l kv e l o c i t y ( b l u e ) i n c o m p a r i s o n t o i n - s i t u m e a s u r e m e n t s ( o r a n g e ) f o r (t o p ) a nd ( b o tt o m ) u s i n ga p e a k find i n ga l go r i t h m . T h e r e d v e r t i c a l b a r s i nd i c a t e t i m e s o f C M E o cc u rr e n ce s a cc o r d i n g t o R i c h a r d s o n a nd C a n e ( ) . SOLA: Paper_Euhforia.tex; 19 August 2019; 2:43; p. 12
UHFORIA background solar wind modeling
Figure 8. a) Drawing of the solar surface features for 2012 May 7 provided by NOAA. TheCH is identified using EUV imagery from spacecraft while the polarity of the CH is obtainedfrom magnetograms. b) Detection of the CH on the same day, by the CHIMERA tool (basedon three wavelengths 211, 193, 171˚A). The image was obtained from Solar Monitor. excluded from the evaluation. In addition, peaks in the in-situ measured so-lar wind speed that could not be related to CHs were also excluded from thestatistical study. We considered observed and modeled solar wind peaks to beassociated, i.e. a hit, if the increase started more or less simultaneously and thepeak was achieved within 2 days after the peak as modeled by EUHFORIA.When the modeled solar wind increase did not have the counterpart in the in-situ observations we considered to have a false alarm, and when the observedfast flow was not reproduced by EUHFORIA we consider to have a miss.The manual identification of the CHs and associated fast flows shows 17 hits,12 misses and 6 false alarms for 2008 and 13 hits, 18 misses and 0 false alarmsfor 2012. We note that these results reveal a significantly smaller number of falsealarms and misses in comparison to the automatic method. This indicates thatthe CH development and its shape has strong influence on the fast solar windspeed profile measured at 1 AU.
In Figures 5 and 6 can be seen that the solar wind modeled by EUHFORIAmatches much better for the interval of the minimum solar activity in 2008. Thismay have several reasons. During the low level of solar activity the magnetic field,the main input for the PFSS extrapolation in EUHFORIA, changes less dynam-ically than during the high level of solar activity, which can result in a morereliable modeling of the solar wind flow. Also, the interplanetary measurementsare not disturbed by transient events which are much less frequent comparedto solar maximum activity, and the solar wind flow is more persistent (Owens et al. , 2013; Temmer, Hinterreiter, and Reiss, 2018).Figure 5a shows for 2008 on average rather good model results of the minimumand maximum solar wind speed, and the majority of fast flows associated withequatorial CHs is well reproduced. However, we also found an exception where
SOLA: Paper_Euhforia.tex; 19 August 2019; 2:43; p. 13 interreiter et al. the in-situ observations show a recurrent fast flow (10 rotations) associated witha well-defined equatorial CH which was modeled by EUHFORIA only at thebeginning of the year 2008. We believe that modeling of the solar wind originat-ing from this particular CH is highly influenced by the CH characteristics anddevelopment in location, size and shape.During the high level of solar activity the magnetic field is very complexand it is known that the amount of low latitude open flux may be significantlyunderestimated by the PFSS model (e.g. MacNeice, Elliott, and Acebal, 2011).Underestimating the open flux leads to significantly lower solar wind speedsmodeled by EUHFORIA. This effect is very strongly pronounced in 2012 (Figure6a). We also note for 2012 the existence of a large number of low latitude CHs sur-rounded by active regions which possibly also influences the model performanceby strongly deviating the magnetic topology from being potential.
During testing of the modeled background solar wind we identified some limi-tations of the present version of EUHFORIA which influences its performance.Herein we identify some of the limitations of the basic setup of the EUHFORIA1.0.4. and a more detailed analysis will be presented in the follow up paper bySamara et al. (2019; in preparation).
Figure 9.
Comparison of model runs with different settings. High/default density: 300 cm − , low/default resolution: 4 ° in longitude and latitude with 256 radial cells (256x30x90).Low density: 150 cm − , high resolution: 2 ° in longitude and latitude with 1024 radial cells(1024x60x180). In order to set up benchmarks for the solar wind modeling with EUHFORIAwe need to understand how different input parameters influence the modeledsolar wind. Figure 9 shows the EUHFORIA model results for the time intervalof several days in March 2008 using different input parameters. We vary theresolution of the heliospheric model and the input density of the fast solar windat the inner boundary compared to the default setting (Section 2.1 herein andSection 2.1.2. in Pomoell and Poedts, 2018). We find that a decrease of the solarwind density by 50% (initial value is 300 cm − at 21.5 R (cid:12) ) induces an increase SOLA: Paper_Euhforia.tex; 19 August 2019; 2:43; p. 14
UHFORIA background solar wind modeling of the modeled solar wind speed from several percent up to 15% (absolute valuedepends on the part of the flow which is considered). Figure 9 also shows acomparison of the default, low resolution runs (angular and radial resolution of4 ° and 256 cells, respectively) and the intermediate resolution runs (2 ° and 512cells, respectively). The higher resolution runs result in an increased solar windspeed (up to about 20%) and in an earlier arrival time of the high speed streamat 1 AU (up to several hours). If we compare the two extreme cases, the defaultEUHFORIA runs i.e., low resolution and high density, and the intermediateresolution and low density runs, we find a shift of the arrival time of the fastflow of about −
12 hours, and a significant increase of the solar wind speed(from about 6% to more than 40%, depending on which part of the fast flowis considered). The obtained results indicate that the quality of the modeledfast solar wind varies a lot depending on the input parameters to the model.We note that when more than one parameter is modified the solar wind speedchanges in a non-linear manner and that the changes strongly depend on theconsidered flow. This brings forward the need for a detailed ensemble parameterstudy which will provide a well-defined benchmark for the solar wind modelingwith EUHFORIA (Samara et al., in preparation).
Comparing CH sizes extracted from EUV observations, and modeled open fluxareas (i.e., CH areas) by PFSS using GONG synoptic magnetograms, shows thaton average CHs are underestimated in the model. It is found that the amountof modeled open flux is lower than actually observed, as well as open flux areasshow up smaller in angular width (Asvestari et al. , 2019). Failure in reliablymodeling open magnetic flux has consequences for a proper solar wind modeling,in particular for the fast solar wind flow originating from CH areas. This willresult not only in an underestimation of the solar wind speed but also mightcause the fast flow to be too narrow, hence, may completely miss the Earth(Section 3.2.3). In a systematic testing it was shown that changing the sourcesurface height (one of the default input parameters to EUHFORIA) significantlyinfluences the modeled open flux and can result even in a shift of the positionof the considered CH (Asvestari et al. , 2019).
While manually associating the observed and modeled solar wind flows (Section3.1.2.), we recognized that the EUHFORIA performance is closely related also tothe size, shape and location of the CHs, sources of the fast flows. The qualitativestudy of the CH characteristics and the quality of the modeled fast solar wind(Section 2.1) shows that for circular and equatorial CHs occurring during thelow level of solar activity, EUHFORIA models well the associated fast flows.However, fast flows associated with narrow CHs elongated in longitude, are rarelyreproduced well by EUHFORIA. In the case of the narrow CHs elongated inlatitude, the modeled solar wind is mostly underestimated, hence, leading toa late arrival at the Earth. And when the solar wind is originating from the
SOLA: Paper_Euhforia.tex; 19 August 2019; 2:43; p. 15 interreiter et al.
Figure 10.
Combined EUHFORIA results in comparison to in-situ data (gray) for the sameinterval as in Figure 9. Earth (blue curve) represents the EUHFORIA output for the Earthlocation. The red curve is an average of combined EUHFORIA results for virtual spacecraft(+4 ° , +8 ° , +12 ° ) above the Earth and the green curve shows the averaged results for virtualspacecraft ( − ° , − ° , − ° ) below the Earth. low/high latitude CHs (greater than ± ° ) and/or the extensions of the polarCHs, it will be rarely reproduced correctly by EUHFORIA. We also noticedthat fast flows associated with patchy CHs, irrespective of their latitudes andlongitudes, are poorly reproduced or not reproduced at all by EUHFORIA.Further on, the fast flows originating from low latitude CHs might pass ’below’or ’above’ the Earth (when the associated CHs are situated at the southern ornorthern solar hemisphere, respectively) and they will not be observed in theEUHFORIA time series output at the Earth (see also Hofmeister et al. , 2018). Inorder to check this hypothesis we have implemented virtual spacecraft around theEarth (separated by 4 ° ranging from − ° to +12 ° in latitude where 0 ° indicatesEarth position) and compared the modeled time series for all these spacecraft.To amplify the effect, the values of time series at +4 ° , +8 ° , +12 ° above the Earthand − ° , − ° , − ° below the Earth were averaged and compared to in-situ data(see Figure 10). We note that the fast flow, starting on March 09, 2008 seem tobe reproduced well by EUHFORIA, by all three time series, i.e. above the Earth,at the Earth and below the Earth. This gives indications on the 3D extent ofthe fast flow that directly impacted the Earth, which is also visible in Figure 1right top panel. The solar wind observed starting from March 19 (Figure 10)originates from rather large low latitude extensions of the southern polar CH.EUHFORIA models at the Earth a somewhat faster solar wind then observedby ACE (blue curve), and significantly faster solar wind passing below the Earth(green curve). In this case the fast flow only glanced the Earth while the mainpart of the fast solar wind passed below the Earth. Studies of the 3D extent ofthe fast flows, using the virtual spacecrafts, is among the main ongoing effortsfor improving our knowledge on the solar wind and solar wind modeling withEUHFORIA (Samara et al., in preparation).
4. Summary and Conclusions
In this paper we present the first results of the solar wind modeling with thenew European model EUHFORIA. For the statistical study we employed the so
SOLA: Paper_Euhforia.tex; 19 August 2019; 2:43; p. 16
UHFORIA background solar wind modeling called basic setup of EUHFORIA 1.0.4. using default input parameters (Section2.1). EUHFORIA currently provides daily modeled results using synoptic GONGmagnetograms. In order to obtain a continuous time series of the backgroundsolar wind parameters, the model outputs from consecutive days have to becombined. We developed a method to derive such a continuous profile from indi-vidual runs taking only the central part of the individual curves and combiningthem using a Gaussian weighting (Section 2.2).We test the quality of the performance of EUHFORIA in solar wind model-ing by selecting two years of different solar activity levels, i.e. 2008 and 2012.The analysis was focused on the comparison of the modeled solar wind for thetwo most important solar wind plasma parameters, i.e. bulk speed and protondensity, and ACE observations (Figures 5 and 6). As a general trend we noticean underestimation of the modeled solar wind speed and an overestimation ofthe modeled density, in comparison with in situ observations by ACE. The solarwind modeled by EUHFORIA matches better for the interval of the minimumsolar activity in 2008, then for the year 2012 when the level of solar activitywas high. We conclude that this result is mostly originating from the betterperformance of the PFSS model (main part of the EUHFORIA’s coronal model)during low level of the solar activity.For defining the association between modeled and observed fast flows we ap-plied an automatic peak finding algorithm (Section 3.1.1). Using this algorithmwe obtain 18 hits, 21 false alarms and 25 misses for 2008 and 14 hits, 7 falsealarms and 24 misses for 2012. As a consequence of the frequently underesti-mated solar wind speed modeled by EUHFORIA, arises the uncertainty in themodeled arrival time of fast streams. Moreover, depending on the CH shape andlocation on the Sun, fast single flows may show multiple wind speed maximawhich restricts the automatic peak finding algorithm in finding the correctlymatching pairs. By visual inspection (Section 3.1.2) we took into account allthese characteristics and assign more reliably the modeled and measured solarwind flow pairs, and obtained better statistics of 7 hits, 6 false alarms, and 12misses for 2008 and 13 hits, 0 false alarms, and 18 misses for 2012.Our statistics show that the quality of the modeled fast solar wind, obtainedusing the basic setup of EUHFORIA and the default input parameters, can bevery variable. In the current study we identified some of the limitations of thissetup. E.g., a higher angular resolution from 4 ° to 2 ° can result in an increase ofthe solar wind speed by up to 20% and with that causes an earlier arrival of thefast solar wind up to several hours. Additionally, as expected high resolution runsshow significantly more structures in the solar wind in comparison to the lowresolution ones. We also tested how the decrease of the fast solar wind densityfrom 300 cm − to 150 cm − influences the modeled solar wind and found thatin the case of the lower input density EUHFORIA will model earlier arrival andlarger amplitudes of the fast solar wind (Section 3.2.1.). When combined, evenonly these two factors can lead to substantial errors in predictions. Detailedanalyzes on such limiting factors are presented in follow-up studies by Asvestari et al. (2019) and Samara et al. (2019; in preparation).The visual inspection of the CHs associated to the fast flows indicates, that theshape and the location of the CHs play an essential role in the model performance SOLA: Paper_Euhforia.tex; 19 August 2019; 2:43; p. 17 interreiter et al. (Section 3.2.3). We found that patchy, elongated and narrow CHs are not wellsimulated by EUHFORIA’s coronal model (i.e., PFSS misses open flux), whichresults in a poor model performance. We also found that the high latitude ( > ° ) CHs, often extensions of polar CHs, may be responsible for EUHFORIAmodeling the fast flow passing above or below the Earth (in a case of CHs onthe northern and southern solar hemisphere, respectively). Therefore, it is veryimportant to have EUHFORIA set up with included virtual spacecraft for allthe future studies of the solar wind modeling by EUHFORIA. This will allowus to estimate the 3D extend of the fast flows and to understand if the fast flowjust missed the Earth, passing below or above it (Section 3.2.3).In the herein presented studies we identified some of the limitations of thepresent version of EUHFORIA 1.0.4. which influences its performance, in partic-ular during the high level of solar activity. We found that the dynamic behaviourof the CHs, together with the complex coronal magnetic field has a major rolein the generation and propagation of the fast solar wind. Due to the complexityof the solar atmosphere modeling of the fast solar wind is a very demandingtask. Herein we presented first attempts to model background solar wind withEUHFORIA, identified some of the limitations of the present setup of the modeland presented first example of the parameter studies. The presented results bringforward the need for a detailed ensemble parameter study which will provide aclear benchmark for the solar wind modeling with EUHFORIA, but which goesbeyond the scope of this paper. The parameter studies, which are presentlyongoing in the framework of the CCSOM project ( ),will help us not only to improve modeling of the solar wind with EUHFORIAbut also to improve EUHFORIA itself. Acknowledgments
J.H. acknowledges the support by the Austrian Science Fund (FWF):P 31265-N27. M.T. acknowledges the support by the FFG/ASAP Program under grant No.859729 (SWAMI). E.A. would like to acknowledge the financial support by the Finnish Academyof Science and Letters via the Postdoc Pool funding. C.S. was funded by the Research Foun-dation – Flanders (FWO) SB PhD fellowship no. 1S42817N. E.A. acknowledges the supportby the Finnish Academy of Science and Letters via a Postdoc Pool grant.
References
Altschuler, M.D., Newkirk, G.: 1969, Magnetic Fields and the Structure of the Solar Corona.I: Methods of Calculating Coronal Fields.
Solar Phys. , 131. DOI . ADS . [altschuler69] Arge, C.N., Odstrcil, D., Pizzo, V.J., Mayer, L.R.: 2003, Improved Method for Specifying SolarWind Speed Near the Sun. In: Velli, M., Bruno, R., Malara, F., Bucci, B. (eds.)
Solar WindTen , American Institute of Physics Conference Series , 190.
DOI . ADS . [arge03] Asvestari, E., Heinemann, S.G., Temmer, M., Pomoell, J., Kilpua, E., Magdalenic, J., Poedts,S.: 2019, Reconstructing coronal hole areas with EUHFORIA and adapted WSA model:optimising the model parameters. arXiv e-prints , arXiv:1907.03337.
ADS . [asvestari19] Bougeret, J.-L., King, J.H., Schwenn, R.: 1984, Solar Radio Burst and In-Situ Determinationof Interplanetary Electron Density.
Solar Phys. (2), 401. DOI . ADS . [bougeret84] Cranmer, S.R., Gibson, S.E., Riley, P.: 2017, Origins of the Ambient Solar Wind: Implicationsfor Space Weather.
Space Sci. Rev. , 1345.
DOI . ADS . [cranmer17] Garton, T.M., Gallagher, P.T., Murray, S.A.: 2018, Automated coronal hole identification viamulti-thermal intensity segmentation.
Journal of Space Weather and Space Climate (27),A02. DOI . ADS . [garton18_1]SOLA: Paper_Euhforia.tex; 19 August 2019; 2:43; p. 18 UHFORIA background solar wind modelingGarton, T.M., Murray, S.A., Gallagher, P.T.: 2018, Expansion of High-speed Solar WindStreams from Coronal Holes through the Inner Heliosphere.
Astrophys. J. Lett. , L12.
DOI . ADS . [garton18_2] Gressl, C., Veronig, A.M., Temmer, M., Odstrˇcil, D., Linker, J.A., Miki´c, Z., Riley, P.: 2014,Comparative Study of MHD Modeling of the Background Solar Wind.
Solar Phys. ,1783.
DOI . ADS . [gressl14] Heinemann, S.G., Temmer, M., Hofmeister, S.J., Veronig, A.M., Vennerstrøm, S.: 2018, Three-phase Evolution of a Coronal Hole. I. 360 ◦ Remote Sensing and In Situ Observations.
Astrophys. J. , 151.
DOI . ADS . [heinemann18] Hofmeister, S.J., Veronig, A., Temmer, M., Vennerstrom, S., Heber, B., Vrˇsnak, B.: 2018, TheDependence of the Peak Velocity of High-Speed Solar Wind Streams as Measured in theEcliptic by ACE and the STEREO satellites on the Area and Co-latitude of Their SolarSource Coronal Holes.
Journal of Geophysical Research (Space Physics) , 1738.
DOI . ADS . [hofmeister18] Jaynes, A.N., Baker, D.N., Singer, H.J., Rodriguez, J.V., Loto’aniu, T.M., Ali, A.F., Elkington,S.R., Li, X., Kanekal, S.G., Fennell, J.F., Li, W., Thorne, R.M., Kletzing, C.A., Spence,H.E., Reeves, G.D.: 2015, Source and seed populations for relativistic electrons: Their rolesin radiation belt changes.
Journal of Geophysical Research (Space Physics) , 7240.
DOI . ADS . [jaynes2015] Jian, L.K., MacNeice, P.J., Taktakishvili, A., Odstrcil, D., Jackson, B., Yu, H.-S., Riley, P.,Sokolov, I.V., Evans, R.M.: 2015, Validation for solar wind prediction at Earth: Comparisonof coronal and heliospheric models installed at the CCMC.
Space Weather , 316. DOI . ADS . [jian15] Karachik, N.V., Pevtsov, A.A.: 2011, Solar Wind and Coronal Bright Points inside CoronalHoles.
Astrophys. J. , 47.
DOI . ADS . [karachik11] Kilpua, E.K.J., Hietala, H., Turner, D.L., Koskinen, H.E.J., Pulkkinen, T.I., Rodriguez, J.V.,Reeves, G.D., Claudepierre, S.G., Spence, H.E.: 2015, Unraveling the drivers of the stormtime radiation belt response.
Geophys. Res. Lett. , 3076. DOI . ADS . [kilpua2015] Kilpua, E.K.J., Balogh, A., von Steiger, R., Liu, Y.D.: 2017, Geoeffective Properties of So-lar Transients and Stream Interaction Regions.
Space Sci. Rev. , 1271.
DOI . ADS . [kilpua2017] Linker, J.A., Miki´c, Z., Biesecker, D.A., Forsyth, R.J., Gibson, S.E., Lazarus, A.J., Lecinski,A., Riley, P., Szabo, A., Thompson, B.J.: 1999, Magnetohydrodynamic modeling of the solarcorona during Whole Sun Month.
J. Geophys. Res. , 9809.
DOI . ADS . [linker99] MacNeice, P.: 2009, Validation of community models: Identifying events in space weather modeltimelines.
Space Weather , S06004. DOI . ADS . [macneice09] MacNeice, P., Elliott, B., Acebal, A.: 2011, Validation of community models: 3. Tracing fieldlines in heliospheric models.
Space Weather , S10003. DOI . ADS . [macneice11] McComas, D.J., Bame, S.J., Barker, P., Feldman, W.C., Phillips, J.L., Riley, P., Griffee,J.W.: 1998, Solar Wind Electron Proton Alpha Monitor (SWEPAM) for the AdvancedComposition Explorer.
Space Sci. Rev. , 563. DOI . ADS . [mccomas98] McGregor, S.L., Hughes, W.J., Arge, C.N., Owens, M.J.: 2008, Analysis of the magnetic fielddiscontinuity at the potential field source surface and Schatten Current Sheet interfacein the Wang-Sheeley-Arge model.
Journal of Geophysical Research (Space Physics) ,A08112.
DOI . ADS . [mcgregor08] McGregor, S.L., Hughes, W.J., Arge, C.N., Owens, M.J., Odstrcil, D.: 2011, The distributionof solar wind speeds during solar minimum: Calibration for numerical solar wind modelingconstraints on the source of the slow solar wind.
Journal of Geophysical Research (SpacePhysics) , A03101.
DOI . ADS . [mcgregor11] Nakagawa, Y., Nozawa, S., Shinbori, A.: 2019, Relationship between the low-latitude coronalhole area, solar wind velocity, and geomagnetic activity during solar cycles 23 and 24.
Earth,Planets, and Space , 24. DOI . ADS . [nagakawa2019] Odstrcil, D., Riley, P., Zhao, X.P.: 2004, Numerical simulation of the 12 May 1997 interplane-tary CME event.
Journal of Geophysical Research (Space Physics) , A02116.
DOI . ADS . [Odstrcil2004] Odstrˇcil, D., Pizzo, V.J.: 1999, Three-dimensional propagation of CMEs in a structured solarwind flow: 1. CME launched within the streamer belt.
J. Geophys. Res. , 483.
DOI . ADS . [odstrcil99] Opitz, A., Karrer, R., Wurz, P., Galvin, A.B., Bochsler, P., Blush, L.M., Daoudi, H., Ellis,L., Farrugia, C.J., Giammanco, C., Kistler, L.M., Klecker, B., Kucharek, H., Lee, M.A.,
SOLA: Paper_Euhforia.tex; 19 August 2019; 2:43; p. 19 interreiter et al.M¨obius, E., Popecki, M., Sigrist, M., Simunac, K., Singer, K., Thompson, B., Wimmer-Schweingruber, R.F.: 2009, Temporal evolution of the solar wind bulk velocity at solarminimum by correlating the stereo a and b plastic measurements.
Solar Physics (1),365.
DOI . https://doi.org/10.1007/s11207-008-9304-7 . [opitz09] Owens, M.J., Spence, H.E., McGregor, S., Hughes, W.J., Quinn, J.M., Arge, C.N., Riley,P., Linker, J., Odstrcil, D.: 2008, Metrics for solar wind prediction models: Comparisonof empirical, hybrid, and physics-based schemes with 8 years of L1 observations.
SpaceWeather , S08001. DOI . ADS . [owens08] Owens, M.J., Challen, R., Methven, J., Henley, E., Jackson, D.R.: 2013, A 27 day persistencemodel of near-Earth solar wind conditions: A long lead-time forecast and a benchmark fordynamical models.
Space Weather , 225. DOI . ADS . [owens13] Paulikas, G.A., Blake, J.B.: 1979, Effects of the solar wind on magnetospheric dynamics:Energetic electrons at the synchronous orbit.
Washington DC American Geophysical UnionGeophysical Monograph Series , 180. DOI . ADS . [paulikas1979] Pomoell, J., Poedts, S.: 2018, EUHFORIA: European heliospheric forecasting informationasset.
Journal of Space Weather and Space Climate (27), A35. DOI . ADS . [pomoell18] Reiss, M.A., Temmer, M., Veronig, A.M., Nikolic, L., Vennerstrom, S., Sch¨ongassner, F.,Hofmeister, S.J.: 2016, Verification of high-speed solar wind stream forecasts usingoperational solar wind models.
Space Weather , 495. DOI . ADS . [reiss16] Richardson, I.G., Cane, H.V.: 2010, Near-Earth Interplanetary Coronal Mass Ejections DuringSolar Cycle 23 (1996 - 2009): Catalog and Summary of Properties.
Solar Phys. , 189.
DOI . ADS . [richardson10] Richardson, I.G., Cane, H.V.: 2012, Near-earth solar wind flows and related geomagneticactivity during more than four solar cycles (1963-2011).
Journal of Space Weather andSpace Climate (27), A02. DOI . ADS . [richardson2012] Rotter, T., Veronig, A.M., Temmer, M., Vrˇsnak, B.: 2012, Relation Between Coronal HoleAreas on the Sun and the Solar Wind Parameters at 1 AU.
Solar Phys. , 793.
DOI . ADS . [rotter12] Schatten, K.H., Wilcox, J.M., Ness, N.F.: 1969, A model of interplanetary and coronal magneticfields.
Solar Phys. , 442. DOI . ADS . [schatten69] Schwenn, R.: 2006, Space Weather: The Solar Perspective.
Living Reviews in Solar Physics ,2. DOI . ADS . [schwenn06] Scolini, C., Verbeke, C., Poedts, S., Chan´e, E., Pomoell, J., Zuccarello, F.P.: 2018, Effect ofthe Initial Shape of Coronal Mass Ejections on 3-D MHD Simulations and GeoeffectivenessPredictions.
Space Weather , 754. DOI . ADS . [scolini18] Temmer, M., Hinterreiter, J., Reiss, M.A.: 2018, Coronal hole evolution from multi-viewpointdata as input for a STEREO solar wind speed persistence model.
Journal of Space Weatherand Space Climate (27), A18. DOI . ADS . [temmer18] Temmer, M., Reiss, M.A., Nikolic, L., Hofmeister, S.J., Veronig, A.M.: 2017, Preconditioningof Interplanetary Space Due to Transient CME Disturbances.
Astrophys. J. (2), 141.
DOI . ADS . [temmer17] Tsurutani, B.T., Gonzalez, W.D.: 1987, The cause of high-intensity long-duration continuousAE activity (HILDCAAs): Interplanetary Alfv´en wave trains.
Planetary and Space Science (4), 405. DOI . ADS . [tsurutani1987] Tsurutani, B.T., Gonzalez, W.D., Gonzalez, A.L.C., Guarnieri, F.L., Gopalswamy, N., Grande,M., Kamide, Y., Kasahara, Y., Lu, G., Mann, I., McPherron, R., Soraas, F., Vasyliunas, V.:2006, Corotating solar wind streams and recurrent geomagnetic activity: A review.
Journalof Geophysical Research (Space Physics) , A07S01.
DOI . ADS . [tsurutani06] Venzmer, M.S., Bothmer, V.: 2018, Solar-wind predictions for the Parker Solar Probe orbit.Near-Sun extrapolations derived from an empirical solar-wind model based on Helios andOMNI observations.
Astron. Astrophys. , A36.
DOI . ADS . [venzmer18] Vrˇsnak, B., Temmer, M., Veronig, A.M.: 2007, Coronal Holes and Solar Wind High-SpeedStreams: I. Forecasting the Solar Wind Parameters.
Solar Phys. , 315.
DOI . ADS . [vrsnak07] Webb, D.F., Howard, T.A.: 2012, Coronal Mass Ejections: Observations.
Living Reviews inSolar Physics , 3. DOI . ADS ..