You are on page 1of 23

Solar Physics (2023) 298:39

https://doi.org/10.1007/s11207-023-02114-9

On the Periodicities in Solar Rotation and Activity


Determined from Sunspot-Group Data and Possible
Connection with Rossby-Like Waves

Domagoj Ruždjak1 · Roman Brajša1 · Ivica Skokić1 · Davor Sudar1 ·


Arnold Hanslmeier2

Received: 28 March 2020 / Accepted: 16 January 2023 / Published online: 10 March 2023
© The Author(s), under exclusive licence to Springer Nature B.V. 2023

Abstract
Analyses of solar rotation and activity, and their variations, can improve the understand-
ing of the formation and interaction of the solar-activity bands and therefore give important
observational constraints on the modelling of the solar dynamo and improve the forecast ca-
pability in solar activity and space weather. Sunspot-position data obtained from the Green-
wich and Debrecen Photoheliographic data for the years 1874 to 2016 were used to calculate
the rotational and meridional velocities of the solar plasma. The velocities were calculated
from daily shifts of sunspot groups. By analysing the rotation-rate residuals, covariance
of meridional velocities and rotation-rate residuals, as well as daily total sunspot number,
taken from WDC-SILSO, for periodicity in the range of periods larger than two days, sig-
nificant periods of their changes were found. It was demonstrated that the frequencies of the
quasi-annual variations of solar activity, with periods from 230 to 400 days, match those of
classical Rossby modes with m ≥ 17. There is no match between the frequencies of Rossby
modes and the peaks in the rotation-rate residual and horizontal component of the Reynolds-
stress-tensor periodograms. These variations are most probably connected with the existence
of active longitudes. It was demonstrated that the algorithms for detecting periodic signals
should be used with caution when analysing complex periodic data.

Keywords Oscillations, Solar · Rotation · Solar cycle · Sunspots, Velocity

1. Introduction

Understanding of the origin and impact of solar-activity changes, which are affecting the
Earth and interplanetary space, is becoming increasingly important as the human presence
in space is progressing. Solar activity, comprising many seemingly different phenomena,
e.g. occurrence of magnetic structures like sunspots and coronal bright points (CBP) or

 D. Ruždjak
rdomagoj@geof.hr

1 Hvar Observatory, Faculty of Geodesy, University of Zagreb, Kačićeva 26, 10000 Zagreb, Croatia
2 Institute of Physics, University of Graz, Universitätplatz 5, 8010 Graz, Austria
39 Page 2 of 23 D. Ruždjak et al.

large energy releases like flares and coronal mass ejections, displays variability on sev-
eral distinct timescales. Those changes include short-term (“Space Weather”) and long-term
(“Space Climate”) variations. The most prominent variation is the 22-year magnetic activity
cycle, whose most obvious manifestation is the 11-year sunspot cycle. Besides this 11-year
Schwabe cycle, both longer- and shorter-term variations are known, such as the ≈ 80 – 90
year Gleissberg cycle, quasi-biennial variations, and the 154-day periodicity (Rieger period).
Extensive reviews on the behaviour of the solar activity are given by Hathaway (2015) and
Usoskin (2017).
Rossby or planetary waves are a specific type of inertial waves occurring naturally in ro-
tating fluids (e.g. Rieutord, 2009) They govern the large-scale dynamics of a rotating sphere,
which is well studied in the Earth’s atmosphere and oceans (see, e.g., review by Dickinson,
1978). Rossby (1939) suggested that the large-scale disturbances in the atmospheric pres-
sure fields could be described by a simple wave equation. These original Rossby waves were
formed when polar air moves toward the Equator while tropical air is moving poleward.
Rossby waves help transfer heat from the tropics toward the Poles and cold air towards the
tropics compensating for the imbalance due to differences in the amounts of solar radiation
received. The slow motion of these waves often results in fairly long, persistent weather
patterns.
Rossby waves owe their existence to the latitudinal variation of the radial component
of the vorticity. Small-amplitude disturbances are assumed around a state of uniform zonal
motion. This is a special case of a barotropic disturbance, conserving absolute vorticity
(Holton, 1992). Applied to atmospheric flow, it takes into account the variability of the
Coriolis parameter while assuming the motion to be two-dimensional. The phase speed is
given by
β
c=U − (1)
k2
where U is the mean westerly flow, β is the derivative of vorticity, also known as the Rossby
parameter, and k 2 is the total wavenumber squared. This equation is known as the Rossby or
planetary-wave formula. Rossby waves are retrograde; they propagate in a counter-rotating
way, to the West on Earth. If β = 0, there will be no Rossby waves. It also means that at
the Equator of any rotating, sphere-like body, including the Earth and the Sun, one will still
have Rossby waves; despite the fact that Coriolis parameter is zero, the Rossby parameter
β > 0.
Differential rotation of the Sun can be explained as generated by rotationally influenced
turbulence in the convective zone. The turbulence leads to the formation of large-scale tur-
bulent fluxes, and the latitudinal flux of the angular momentum is described by the hori-
zontal component of the Reynolds-stress tensor. Reynolds stress is the main contributor to
the transport of angular momentum towards the solar Equator that maintains the observed
solar differential rotation (Sudar et al., 2014, 2017). Residual velocities larger then average
are positively correlated with meridional motions towards the Equator. Since Rossby waves
contribute a significant part of the Sun’s large-scale kinetic energy (Löptien et al., 2018),
they may play an important role in the dynamics of the solar interior. They might have
influence on the Sun’s differential rotation and magnetic dynamo, the main driver of the
solar activity. The physical processes underlying the complex magnetohydrodynamical sys-
tem are not fully understood, and the origin and effects of Rossby waves are still unknown.
Their formation might be caused by solar rotation, magnetic fields, and other factors, such
as g-mode oscillations (Saio, 2018), and they could be different in various layers of the Sun
(Khomenko and Collados, 2015).
Periodicities in Solar Rotation and Activity and Connection with Rossby Waves Page 3 of 23 39

The recent discovery of Rossby waves on the Sun (Löptien et al., 2018) provided another
possible explanation of the observed solar-activity changes. Rossby waves were hypothe-
sized as the possible cause of solar-activity variations even before their discovery (Gilman,
1968, 1969a,b; Zaqarashvili et al., 2010a,b, 2015). McIntosh et al. (2015) found quasi-
annual variability in flare, CME, and sunspot occurrence, as well as the radiative and par-
ticulate environment of the heliosphere, while analysing variations of solar magnetism dur-
ing Solar Cycle 23. They inferred that this was a secondary variability driven by surges
of magnetism from activity bands of the solar cycle. More recently, while studying the
evolution of coronal bright points from the Solar Dynamics Observatory (SDO) and So-
lar Terrestrial Relations Observatory (STEREO) spacecraft, McIntosh et al. (2017) found
that the global-scale bands of magnetized activity slowly meander westwards in longitude
and display Rossby-wave-like behaviour. These were interpreted as evidence of “Rossby-
like waves”, although the directions of the phase and group velocities are the opposite of
what is expected for classical Rossby waves (westward on the Sun is the prograde direc-
tion). Therefore, one may ask whether there is a connection between the internal Rossby
waves and the large-scale travelling patterns of magnetic bright points in the Sun’s corona
reported by McIntosh et al. (2017). Finally, Löptien et al. (2018) unambiguously detected
and characterized vorticity waves propagating retrograde in the shallow subsurface layers
of the Sun, with the dispersion relation of sectoral Rossby waves and well defined mode
frequencies corresponding to periods larger than twice the solar rotation period. The waves
have lifetimes of several months and nearly as much vorticity as the convection at the same
scales; thus, they are an essential component of solar dynamics. Liang et al. (2019) con-
firmed the existence of global equatorial Rossby waves in the solar interior over the past
two solar cycles by time–distance helioseismology.
In this article, we investigate the nature of the solar-rotation and -activity variations, and
examine the possible connection of the solar mid-term variations with the recently discov-
ered solar Rossby waves in the solar interior. Such studies are important because they can
give important observational constraints on the modelling of the solar interior and improve
the forecast capability in solar activity and space weather. The purpose of our study is man-
ifold:
i) To identify significant periods of changes of solar rotation and activity. To this end,
we performed the analysis of long data series spanning more than a century. The data
obtained using sunspots and sunspot groups are analysed for periodicity in the range of
frequencies from 0 to 0.5 day−1 , which corresponds to the periods larger than 2 days.
This is described in Sections 3.1 and 3.2.
ii) To emphasize that the algorithms used for determining and characterizing periodic sig-
nals in astronomical data should be used with caution, because the periodic signature
from the sampling frequency of the data is coupled with the periodic nature of the
source, which might lead to determination of false periods (cf. Section 3.1.2).
iii) To examine which frequencies are present on which time intervals and how they corre-
late to the solar-activity level. To this end, we calculated the wavelet power spectra of
all quantities and compared them to solar-activity level (cf. Sections 3.1.3 and 3.2.3).
iv) To discuss (in Section 3.3) how those changes could be explained, especially in the
framework of the recent discovery of Rossby waves on the Sun.

2. Data and Analysis


The daily total sunspot number [Sn ] was used as a proxy for the solar activity. The data
were taken from WDC-SILSO, Royal Observatory of Belgium, Brussels (SILSO World
39 Page 4 of 23 D. Ruždjak et al.

Data Center, 1874 – 2016) The version 2.0 of the data containing a new, recently entirely
revised data series, available since 1 July 2015, was used.
The sunspot-group data were taken from the Greenwich Photoheliographic Results
1874 – 1976 (GPR) and its continuation the Debrecen Photoheliographic Results (DPD)
(Baranyi, Győri, and Ludmány, 2016; Győri, Ludmány, and Baranyi, 2017) sunspot
databases (1977 – 2016). Our own GPR digital dataset was used and DPD were downloaded
from their website. The catalogue data include the reference number of the sunspot groups
and, among others, the following daily parameters: the time of observation, heliographic
coordinates of the group centers, mean umbral and whole spot areas, and the classification
of sunspot groups.
The rotation-rate residuals [ω] and their product with meridional velocity, i.e. hori-
zontal component of the Reynolds-stress tensor [ω vm ] (Sudar et al., 2014, 2017), were
used as proxies for the solar differential rotation. Daily positions of the sunspot groups have
been used to obtain the rotational and meridional velocities. For the groups for which the
central meridian distance (CMD) was less than 58◦ , which corresponds to about 0.85 of
the projected solar radius (see Balthasar, Vazquez, and Wöhl, 1986), the rotation rate was
calculated by the division of CMD differences by the elapsed time and the meridional ve-
locity by the division of latitude differences by elapsed time. The obtained synodic-rotation
velocities were then transformed to sidereal ones by the procedure described by Roša et al.
(1995), Brajša et al. (2002), and Skokić et al. (2014). Finally, the sidereal-rotation velocities
less than 8◦ day−1 or exceeding 19◦ day−1 have been discarded as outliers. Rotation-rate
residuals were calculated by subtracting the individual rotation velocities from the average
rotation profile of all data:

ω(b) = A + B sin2 b, (2)

where A and B are the differential-rotation parameters in ◦ day−1 and b is the heliographic
latitude in ◦ . To avoid false meridional flows (Olemskoy and Kitchatinov, 2005), to each
calculated rotational and meridional velocity we assigned the latitude of first measurement
from the pair used in calculation. More details on data reduction are given in the Appendix.
Finally, to obtain the dataset equivalent to the sunspot-number dataset, daily mean values,
obtained by taking a simple arithmetic mean, were taken as the measure for the solar rotation
of a given day. If on a given day no sunspots were present within 58◦ from central meridian,
the appropriate value (“NaN” or zero) was used to preserve equidistant dataset. In this way,
all datasets have values for each day, since the total sunspot number is equal to zero on days
without sunspots.
The obtained data were analysed for periodicity in the range (2 − ∞), i.e. periods larger
than two days using Fourier transformation, Lomb–Scargle periodogram, and wavelet anal-
ysis. The Fourier analysis was performed using the Period04 (Lenz and Breger, 2014) soft-
ware, while the Lomb–Scargle periodograms and wavelets were computed using the Matlab
and Python programming languages, respectively.

3. Results and Discussion

The periods of solar-rotation and activity variations can be divided in groups according to
their timescale. Unfortunately, there is no unique nomenclature in the previous works, e.g.
the changes that take place on the timescales of one or two years were called: “mid-term
quasi-periodicities” (Mursula, Zieger, and Vilppola, 2003), “quasi-biennial oscillations”
Periodicities in Solar Rotation and Activity and Connection with Rossby Waves Page 5 of 23 39

(Bazilevskaya et al., 2014), and more recently “quasi-annual variations” (McIntosh et al.,
2015). We will use the following division: Long-term variations comprising Gleissberg and
Schwabe cycles and changes within the cycle with periods larger than ≈ 2.5 years. Mid-term
variations with periods ranging from 0.5 – 2.5 years, which encompass quasi-annual and -
biennial variations. Short-term variability including from Rieger period to periods compara-
ble to the solar rotation to periods of several days. Here the millennial and rapid changes with
periods in the range of one day or shorter, are omitted. However, they cannot be analysed by
our dataset anyway due to its limited time span and the Nyquist frequency, corresponding to
a two-day period.
Analyses of the data by using the Fourier transformation and Lomb–Scargle periodogram
are giving essentially the same result. An attempt to model solar variations by least-squares
fitting of the linear combination of several sine waves to the data, i.e. removing the Gleiss-
berg or Schwabe cycle variations from the data (using Period04), did not yield significantly
different results. Although the shape and importance of some peaks is changing, there is no
significant difference between the results of the analyses undertaken.
Further, we would like to emphasize that although the word “period” is used when de-
scribing the periodograms, one should bear in mind that the nature of the described changes
might not be periodic but quasi-periodic, cyclic, or even stochastic, i.e. the amplitude and
period are changing from cycle to cycle and/or the variations are not always present. An-
other important note is that all peaks in periodograms are not independent and some might
be harmonics or a consequence of aliasing or spectral leaking and therefore probably do not
necessarily represent the real changes. The main question is: Which of the periods found are
independent of each other and what physical processes on the Sun are causing the observed
variations?

3.1. Long-Term Variations

The Lomb–Scargle periodograms of the daily total sunspot number, daily values of rotation-
rate residual, and horizontal component of the Reynolds-stress tensor are shown in Figure 1.
Dotted vertical lines denote 33-, 22-, and 11-year periods and the gray shaded area denotes
the frequency range that corresponds to periods between 312 and 400 days (centered at 1-
year period), while horizontal lines denote False Alarm Probability (FAP) levels: 0.1% solid
lines, 10% dashed lines. It can be seen that long-term variability is significant in solar activity
and rotation-rate residuals. The strongest signal in the sunspot-number data is the 11-year
cycle, followed by the 145-year peak, which represents either the double Gleissberg cycle or
the long-term secular changes. The peak is very broad due to the shortness of the observing
interval when compared to the period and does not permit precise period determination.
Similarly, the rotation-rate residual spectra have the strongest peak at 35 years, which is
roughly three times the Schwabe 11-year period, followed by the peaks at 188 and 71 years.
In addition, the rotational velocities show variations with periods of 27.1, 7.8, and 11.6 years
and the sunspot number changes with periods of 6.4, 5.2, 4, and 3.2 years. The complete
list of significant periods for sunspot number, rotation rate, and Reynolds stress is given in
Table 1.
Contrary to the solar activity and rotation rate, in the horizontal component of the
Reynolds-stress periodogram, statistically most significant are secular (P > 100 years)
changes, followed by the 11- and 22-year signal. Besides, only the peak corresponding to
90-year period has FAP less than 0.1%. This result is not surprising. While the long-term
variations of solar activity and rotation are well documented (see, e.g., Brajša, Ruždjak, and
Wöhl, 2006; Ruždjak et al., 2017, and references therein), only very weak changes of the so-
lar differential rotation were previously found (Ruždjak et al., 2017, and references therein).
39 Page 6 of 23 D. Ruždjak et al.

Figure 1 Lomb–Scargle power


spectral density for daily total
sunspot number (top), daily
values of rotation-rate residual
(middle), and daily values of
horizontal component of
Reynolds-stress tensor (bottom).
Gray shaded area is centered at
the frequency corresponding to
1-year period and encompasses
the 312- to 400-day period range.
The horizontal lines denote the
False Alarm Probability (FAP)
levels. Solid and dashed lines
denote 0.1% and 10% FAP level,
respectively. Vertical lines and
arrows are described in the text.

If the Reynolds stress is the mechanism that maintains the solar differential rotation, and
no prominent long-term changes of the differential rotation are observed, then one should
not expect prominent long-term variations of Reynolds stress. In any case, changes not as
prominent as the variation of rotation rate or solar activity are expected. However, this might
be at least partly due to poorer signal-to-noise ratio in Reynolds stress data and is discussed
in more detail in the Appendix.

3.1.1. Gleissberg and Schwabe Cycle

The periods representing Gleissberg and Schwabe cycles are well known and found by many
authors. The value of the Gleissberg-cycle length of 80 – 90 years and the mean value of 11
years for the Schwabe cycle (Hathaway, 2015, and references therein) correspond well to
the ones found here. A relatively short data span, when compared to the Gleissberg-cycle
length, does not allow precise period determination, but 90 ± 24 and 71 ± 12 years found in
the Reynolds stress and rotation-rate residual correspond well to the previously found values
and the longer periods found are roughly twice the Gleissberg period.
Hovewer, the 71 ± 12 years period is exactly double the 35 ± 3 years period, the strongest
peak in rotation-rate residuals, and we cannot exclude that it is a harmonic of the Schwabe-
cycle period. The long period (140 – 190 years) might also represent secular changes whose
period cannot be determined by our dataset. We can only detect with certainty the Schwabe
cycle upon which some changes with longer timescale of unknown character (cyclic or
stochastic) are superimposed. From inspection of the position of peaks in periodograms,
it might seem that the mean cycle obtained from rotation is slightly larger than the one from
sunspot number, but the value of 10.8 ± 0.6 years found from sunspot-number data does not
differ significantly from 11.6 ± 0.9 years found in the rotation rate, or 11.2 ± 0.9 years and
12.4 ± 1.1 years found in the Reynolds-stress periodogram.
Periodicities in Solar Rotation and Activity and Connection with Rossby Waves Page 7 of 23 39

Table 1 Periods of solar-rotation and activity variations found in this work. The periods are grouped in
groups according to period length. All peaks with FAP less than 10% are listed, those with FAP less than 1%
boldfaced. Additionally, peaks in the period range 10 to 2000 days in solar rotation rate with FAP less than
50% are shown in italics. In first two columns, the azimuthal wavenumber m and corresponding period Pm
for m =  Rossby mode in the frame rotating at the surface equatorial angular velocity  = 1/25.65 days
(453.1 nHz) are given. The errors of the periods were calculated from estimated widths at half maximum [f ]
of the corresponding peaks as P = f /f 2 . SN stands for daily total sunspot number, RR for rotation-rate
residual and RS for horizontal component of Reynolds stress.

m Pm Period (SN) (RR) (RS) Remark

Periods in years
145±153 188±260 71±12 90±24 Gleissberg cycle
44±12 33±7 22±5 47±25 35±3 Schwabe cycle
20±2 27±2 20±7 ”
14±1 13±1 12±1 ”
10.8±0.6, 5.2±0.2 11.6±0.9 11.2±0.9 ”
8.7±0.6 8.3±0.5 Influenced by
7.8±0.5 7.3±0.5 Schwabe cycle
6.4±0.2 Aliasing and
4.8±0.2 annual and
4.0±0.1 3.91±0.08 4.0±0.2 biennial
3.8±0.1 3.7±0.2 harmonics
3.59±0.08 3.5±0.1 ”
3.2±0.1 ”
2.62±0.07 2.50±0.08 2.40±0.03 QBO +
2.30±0.07 2.11±0.05 harmonics
2.00±0.07 1.96±0.03 1.99±0.03 QBO +
1.88±0.02 1.85±0.02 harmonics
1.76±0.02 ”
34 1.22 1.34±0.02
33 1.19 1.29±0.02
32 1.15 1.22±0.02
31 1.12 1.15±0.01

Periods in days
30 395.9 399±3
29 383.2 389±3
28 370.4 379±3
27 357.6 372±3
26 344.9 366±3 368±3
25 332.1 349±3
24 319.3 335±3
23 306.5 321±3
22 293.8 280±3
21 280.9 274±2
20 268.2 252±2
19 255.4 247±2
18 242.7 239±2
17 229.9 236±2
12 166.0 157.8±0.5
11 153.3 156.2±0.5 Rieger period
10 140.5 134.2±0.4
7 102.2 97.1±0.2
4 63.9 28.03±0.02, 28.15±0.02 61.35±0.08 Solar rotation
1 25.5 28.00±0.02, 27.96±0.02 20.53±0.06 18.57±0.01 ”
27.80±0.02 27.77±0.02 15.98±0.03 ”
27.50±0.02, 27.46±0.03 14.18±0.03 ”
27.11±0.01, 27.06±0.01 ”
26.95±0.02, 26.85±0.02 ”
26.75±0.01, 26.72±0.01 ”
26.27±0.01 25.89±0.01 ”
39 Page 8 of 23 D. Ruždjak et al.

3.1.2. Periods within the Cycle

The period groups as defined here and in some previously published works are not inde-
pendent, as periods in different groups might be caused by the same physical processes on
the Sun. Annual and biennial changes are hardly independent from each other, in the sense
that the annual frequency is twice the biennial one. For instance, in the sunspot-number
periodogram, strong peaks at 3.2, 2.1 years, and 399 days are observed (3.2 ± 0.1 years,
2.11 ± 0.05 years and 1.09 ± 0.08 years), and we cannot decide which of them, if any,
represent real variations and which are the harmonics. Here, when referring to annual and
biennial periods, we mean the periods from 0.5 to 2.5 years or periods connected (by being
a harmonic) to such period. Therefore, the periods in the range 2.5 – 10 years, (here denoted
as periods within the cycle) were found before by many authors. For instance, Brajša, Ružd-
jak, and Wöhl (2006) found 5.2 and 3.5 year periods while analysing rotation-rate residual
changes. Most recently, Deng et al. (2020) found 11.13, 9.53, 6.13, and 3.25 years as sig-
nificant periods of coronal rotation for the 1939 – 2019 time interval. Those periods are very
similar to the ones found here, and it is reasonable to assume that they are of the same ori-
gin. In this work, most, if not all, of those periods found in solar rotation are artifacts of the
Schwabe solar cycle. To show this, the spectral window of daily number of sunspots and
rotation-rate residual calculated with Period04 is shown in Figure 2c and d. The horizontal
component of the Reynolds-stress window is identical to the rotation-rate residual window,
since both datasets have exactly the same sampling. While the amplitude of the side peaks
is exponentially decreasing with frequency in the sunspot-number spectral window, it has
prominent peaks at 10.5 (0.00026 day−1 ) and 5.2 (0.00052 day−1 ) years for the rotation-rate
residual and horizontal component of the Reynolds stress. This is obviously a consequence
of the Schwabe cycle, since the peaks are at the frequencies corresponding to its cycle length
and its second harmonic. This is entirely a consequence of data sampling and is due to the
days when no sunspots were observed on the Sun, which repeats roughly every 11 years. Al-
though not relevant for the current discussion, the peak corresponding to the solar rotation at
0.037 day−1 (27 days) is also visible in the rotation-rate residual and horizontal component
of the Reynolds-stress spectral window.
Now, since for each analysed frequency, for which the Fourier transform is calculated,
instead of convolution with a delta function, the convolution with spectral window is calcu-
lated, each significant peak might have side peaks (aliases) at frequencies f ± f , where
f is the frequency of the peak in the spectral window. As can be seen in the case of 11-
year peak in the sunspot-number periodogram, which is plotted on the same logarithmic
scale as the spectral window, on both sides of the main peak the attenuated side peaks are
seen with amplitude of about 10% of the main-peak amplitude. The same can be seen in
the rotation-rate periodogram on the 7.8-year (0.00035 cycles day−1 ) peak, but with the side
peaks at 50% of the main-peak amplitude. This is due to aliasing. If the most significant
peaks correspond to the first two harmonics of the Gleissberg cycle and the third harmonic
of the Schwabe cycle at 3.9, 1.8, and 8.3 ×10−5 day−1 (70, 140, and 33 years, as is the
case in our analysis of rotation rate) then the side peaks due to 10.5 and 5.2 peaks in the
spectral window will be at frequencies corresponding to: 9.7, 9.1, 7.9, 5.0, 4.9, and 4.4 years
(denoted by arrows in the periodogram). Consequently, the amplitude of the fourth strongest
peak in the rotation-rate residual spectrum (corresponding to a period of 7.8 years) is at least
mostly, if not completely, a consequence of the 10.5-year alias of the third harmonic of the
Schwabe cycle. The aliases of the 11-year period would be located at 5.4 and 3.5 years.
Taking into account also the interference between peaks, one can imagine that the presence
of the 11-year side peak in the spectral window results in very complicated spectra, even
Periodicities in Solar Rotation and Activity and Connection with Rossby Waves Page 9 of 23 39

Figure 2 (a) Total daily sunspot-number periodogram in the frequency range corresponding to solar rota-
tion. (b) Rotation-rate residual and horizontal component of Reynolds-stress tensor spectral window in the
frequency range corresponding to solar rotation. (c) Spectral window for the total daily sunspot-number data
sampling. (d) Spectral window for the rotation-rate residual and horizontal component of Reynolds-stress
tensor data sampling.

if only the Schwabe cycle represents a real signal. This discussion does not apply for daily
sunspot number, but the 5.2-year is a harmonic of the Schwabe cycle, while other peaks
might be also connected with the annual quasi-periodicity.
Another interesting fact is that the beat period of the two most significant peaks con-
nected to the solar rotation (shown in the Figure 2a) is about 13 years, fairly close to the
Schwabe-cycle length. The peak at this frequency is visible in the periodogram and is de-
noted with the arrow in the top panel of Figure 1. Although its existence can be entirely due
to a frequency leakage from the 11-year peak, it is mentioned here to illustrate how difficult
it is to determine what are the real periods and nature of the changes, even without taking
into account that the analysed changes are not strictly periodic but cyclic and that the cycle
length (period) is changing from cycle to cycle.
Based on the previous discussion, we conclude that all of the periods in the range of
2.5 – 10 years (denoted by rectangle in Figure 1) found here and previously by other authors
are strongly influenced by the artifacts of the Schwabe cycle and harmonics of other frequen-
cies, and therefore they are unreliable and most probably do not represent real variations of
solar rotation or activity.

3.1.3. Changes of the Long-Term Variability and Connection to the Solar Activity

In order to examine when some periods of variations in the solar activity and rotation are
present, a wavelet analysis of the data was performed. In Figure 3 the wavelet power spec-
trum of rotation-rate residuals is shown with the 13-month smoothed monthly total sunspot
number. On the long-period side in the figure, two stripes at about 70 and 150 years can
39 Page 10 of 23 D. Ruždjak et al.

Figure 3 Wavelet power spectrum of rotation-rate residual using Morlet wavelet functions plotted together
with the 13-month smoothed sunspot number. The solid line on wavelet spectra denotes the “cone of influ-
ence” (COI) line, outside of which the edge effects become important. The horizontal dashed lines represent
periods of 1 and 32 years and vertical ones epochs of solar minima.

be noted, representing the Gleissberg cycle. The strongest signal from the Schwabe cycle is
at about 32 years, and it can be noted that the period is decreasing during the first half of
the 20th century attaining its current value around 1960, coinciding with the modern solar
maximum. Unfortunately, most of the described variability is outside the “cone of influ-
ence” (COI) lines, where the signal is decreased by the influence of data-interval edges,
but the “knee” of the change around Solar Cycle 17 and 18 is not influenced by the edges.
This implies that as the level of the solar activity is rising, i.e. the global magnetic field
is strengthening, the individual cycles are getting shorter, attaining a minimal value during
Solar Cycle 19, coincident with the grand maximum of solar activity, and have stayed at
this level since. With the error of the period determination of about three years for the third
harmonic of the Schwabe cycle this result is statistically significant only outside the COI
line and not significant inside the COI line, but the trend can be clearly seen on the figure.
Another interesting feature seen in Figure 3 is that, with the exception of Solar Cycles 12
and 21, the biennial, annual, and shorter variability is confined to temporal intervals around
the minimum of activity and suppressed during the modern solar maximum.

3.2. Mid- and Short-Term Variations

3.2.1. Solar Rotation

To avoid any possible confusion, we start this section by stressing that in dealing with solar
rotation, sidereal velocities and periods are always used. However, when one is examining
repeating patterns in sunspot number, rotation-rate residual, and Reynolds stress due to rota-
Periodicities in Solar Rotation and Activity and Connection with Rossby Waves Page 11 of 23 39

tion of the Sun, the synodic-rotation period is appropriate since the observations from Earth
were used.
On the short-period side, there is a notable peak at about 27 days in the daily sunspot-
number data (Figure 1, top), which is the reflection of the Sun’s rotation period. The solar-
rotation period is detected in the sunspot-number periodogram as six significant peaks
(FAP< 0.1%) (see Figure 2a), with the two strongest located at 27.3 and 27.4 days. On
the other hand, in rotation rate and Reynolds stress only several sharp peaks, attaining
power over 50% FAP level, are seen at frequencies corresponding to 10 to 300 days. The
strongest being at 18.6 days, attaining over 90% FAP level in the horizontal component of
the Reynolds-stress periodogram. This is mainly a consequence of the solar differential rota-
tion and data preprocessing. Some sunspot groups (or active regions, more generally) can be
present on the Sun for several rotations, giving rise to the patterns that are repeating with the
rotation period of the Sun. But sunspots are arising on activity bands, which move toward
the solar Equator during the course of the solar cycle, and the observed rotation period is
shortening as sunspots are generated at lower and lower latitudes. In this case, the multiple
peaks are the consequence of both the change of repeating patterns and the change of the
rotation-period length during the solar cycle. On the other hand, during data pre-processing
for rotation-rate residual, the latitudinally dependent mean differential rotation profile was
subtracted from the data. In this way, the influence of the rotation-period change during the
solar cycle was canceled. However, a similar structure of periods to the one observed for
solar rotation in the sunspot-number periodogram, showing several peaks near the rotation
period, can be observed in the spectral window for the solar-rotation and Reynolds-stress
data (Figure 2b and d). Again, this is due to the spotless periods when there are no data
in rotation-rate residual and Reynolds-stress data sets and the “NaN” value was used. The
sunspot number on those days is equal to zero, which smoothly encloses the daily total
sunspot-number dataset. Similarly, as before, a spotless disc is giving rise to the patterns
that are repeating with the rotation period of the Sun. Also, the period range of peaks of
26.1 – 27.3 days in the spectral window, with a highest peak at 26.24 days, is slightly lower
than the one in the daily-sunspot-number periodogram: 26.7 – 28.0 days, with the highest
peaks at 27.3 and 27.4 days. This is due to the spotless days being present around the mini-
mum of activity, when sunspots are mainly near the Equator, therefore describing equatorial
rotation, while the peaks in the sunspot-number periodogram describe mean solar rotation.
The values found here are in a good agreement with the synodic equatorial rotation period
of 26.24 days and Carrington rotation of 27.2753 days (Stix, 2002).

3.2.2. Quasi-Biennial Variations

To examine more closely the complex periodograms, the range of periods from 10 to 2000
days is shown in Figure 4. The periodogram of daily sunspot number is very complex with
many significant peaks in the range 100 to 1000 days and three peaks in the vicinity of
the one-year period with the FAP less than 0.1%, namely 398.9, 378.8, and 365.5 days. In
accordance with the finding that splitting of periods due to the change of rotation period
during the solar cycle was removed by data pre-processing, the periodogram of the rotation-
rate residual looks simpler and statistically the most significant peak is at 368 days. There are
several peaks for which the FAP is less than 50%, namely 4.8, 4, 2.4, and 2 years and 61.3,
28.7, 20.5, 16.0, and 14.2 days, from left to right. There are only three peaks in the horizontal
component of the Reynolds-stress periodogram that attain power larger than 50% FAP level:
5.3 ± 0.5 years, 4.6 ± 0.4 years, and 18.57 ± 0.01 days. All peaks in the periodograms are
not independent from each other, for instance the peaks at 4 and 2 years are most probably
39 Page 12 of 23 D. Ruždjak et al.

Figure 4 Lomb–Scargle power


spectral density for daily total
sunspot number (top), daily
values of rotation-rate residual
(middle), and daily values of
horizontal component of
Reynolds-stress tensor (bottom)
in the range of periods 10 to 2000
days. The black lines denote
harmonics of the frequency
corresponding to one-year period
and are centered at frequencies
corresponding to periods of 4, 2,
1, and 0.5 years. The horizontal
lines denote the False Alarm
Probability (FAP) levels. Solid,
dashed, and dotted lines denote
0.1%, 10%, and 50% FAP level,
respectively.

harmonics of the one-year period as well as 2.4- and 4.8-year periods, which are each other’s
harmonics. The rotation-rate periodogram is more complicated for periods larger than about
two years; this is not due to real variations, but due to this range being influenced by aliasing
from the Schwabe and Gleissberg cycles.

3.2.3. Changes of Mid-Term Variations and Connection to the Solar Activity

Before proceeding with discussion of possible causes of periodicities found in the solar
activity and rotation, let us examine when some periods of changes in the solar activity and
rotation are present. In Figure 5 the wavelet power spectrum of the daily sunspot number,
rotation-rate residual, and the horizontal component of the Reynolds stress in the range of
periods up to 5 years is shown with the 13-month smoothed monthly total sunspot number. It
can be seen that the behaviour of daily sunspot number is opposite to that of the rotation-rate
residual, i.e. the variations are not present in the intervals around the minimum of activity.
The absence of variations of sunspot number in the cycle minima is not surprising, since
one cannot expect to find variations of something that is not present, i.e. the sunspot number
becomes too small to detect its changes. The variations of the horizontal component of the
Reynolds-stress tensor are not as confined to the solar-minimum epochs as the rotation-
rate residual, especially during Solar Cycles 21 – 24, but they are suppressed and confined
to minimum instances, just before and during the modern solar grand maximum. It might
seem that the main period of the quasi-annual (/biennial) oscillation of the sunspot number
is changing regularly during the first half of the 20th century. The period changes from 0.6
years in Solar Cycle 14 to about 2 years in Solar Cycle 15 and gradually again to less than
a year in Solar Cycle 18. However, those variations are too large to be the consequence
of changes due to rise of solar activity to the modern maximum, and they are more likely
Periodicities in Solar Rotation and Activity and Connection with Rossby Waves Page 13 of 23 39

Figure 5 Wavelet power spectrum of daily total sunspot number (b), rotation-rate residual (c), and horizontal
component of Reynolds stress (d) using Morlet wavelet functions in the range of periods from 0 to 5 years,
plotted together with the 13 month smoothed sunspot number (a). As before, the solid lines on wavelet spectra
denote the COI lines. The horizontal dashed line denotes the one-year period and vertical ones denote epochs
of solar minima.

to represent switching between different harmonics of the same quasi-biennial period or


between different modes of oscillation.
39 Page 14 of 23 D. Ruždjak et al.

3.3. On Possible Causes of the Observed Mid-Term Variations

Now when we mention oscillations, let us consider that the observed periods of changes are
caused by oscillations in the solar interior. Löptien et al. (2018) discovered and character-
ized 13 modes of Rossby waves with wavenumbers below 15, at which point they become
indistinguishable from convection, with a range of periods from 45 to 246 days. Liang et al.
(2019) confirmed the existence of global equatorial Rossby waves over the past two solar
cycles, with the modes having lifetimes of a few months to more than a year, in agreement
with the distribution of periods in Figure 5. Rossby waves occur naturally in rotating fluids
and govern the large-scale dynamic of rotating sphere. Their dispersion relation is given by
2m
ω= , (3)
( + 1)
where  is the integral spherical harmonic degree, m is the azimuthal wave-number, and
 is the rotation velocity of the uniformly rotating sphere. To see how this simple model
fits our results, the calculated periods of Rossby-wave modes are given in Table 1. Only
 = m modes are calculated, since those are the only ones discovered so far, using the solar
rotation period of −1 = 25.65 days, which is the mean solar-surface equatorial rotation
for 2010 – 2016 from SDO/HMI observations (Löptien et al., 2018). The frequencies do not
match on the low-period side, while on the high-period side the matching is better. This
does not mean that we can eliminate the Rossby waves as a cause of variability in the solar
activity and rotation, especially as a driver of quasi-annual variations. It should be pointed
out that the Sun is not a rigidly rotating sphere and magnetic fields play a significant role in
the fluid dynamics, and, therefore, it is not straightforward how the Rossby waves would be
influencing sunspots. This might result in significant deviations of frequencies from the ones
predicted by a simple model, but the poor overlap of the periods found with the frequencies
of Rossby-wave modes for the low- modes makes the Rossby waves less convincing as
the explanation of all of the observed periodicities. It is also worth mentioning that Liang
et al. (2019) obtained periods about twice as large for 11 ≤ m ≤ 15 modes by analysing the
SOHO/MDI and SDO/HMI data, when compared to Löptien et al. (2018).
Another possible explanation of the periods, found in the range of 10 to 300 days, is
that they are caused by the solar rotation itself. In the rotation-rate-residual periodogram
the peaks at 61.4, 28.7, and 20.5 days can be noted. Their frequency ratio is roughly 1:2:3,
and 28.7 days corresponds to the solar rotation period, but the ratio is not exactly right and
28.7 days is slightly too large to be straightforwardly connected with the sunspot rotation.
However, if it is assumed that various manifestations of solar activity are non-axisymmetric
and mainly occur in several preferred longitude ranges, then different periods could be ob-
tained. Analysing the occurrence of solar X-ray flares (Zhang et al., 2011a) and sunspots
(Berdyugina and Usoskin, 2003; Usoskin, Berdyugina, and Poutanen, 2005; Zhang et al.,
2011b), it was found that the longitude distribution shows two persistent preferred longi-
tudes separated by about 180 degrees whose strength alternates in time. Let us imagine that
there are two fixed points separated by 180 degrees on the solar surface, at which a certain
number of sunspots (or flares) is present each day and those configurations can survive sev-
eral solar rotations. Due to the solar rotation, one would observe repeating patterns at half
the rotation period. Therefore, the expected rotation signal would be at half the rotational
period. There are two peaks in that range in the rotation-residual periodogram at 14.2 and
16.0 days. Both might represent solar rotation, since the active longitudes are not separated
by exactly 180 degrees and their strength (and possibly position) changes with time. The
14-day period can further be connected with the Rieger period (154 = 14 × 11) and further
Periodicities in Solar Rotation and Activity and Connection with Rossby Waves Page 15 of 23 39

Figure 6 Measured peak frequencies in the daily total sunspot-number power spectrum and theoretical mode
frequencies (curve). Circles denote the FAP < 1% peaks. The frequency errors are in the order of 1% and can
be found in Table 1.

with the 2.4-year period (14 × 7 × 9). Even the one-year period can be connected with the
14-day period (364 = 14 × 26). In this case, along with the assumption that multiple peaks
in sunspot number are caused by the change of rotation period due to the movement of activ-
ity bands during the solar cycle, all observed changes, besides the Gleissberg and Schwabe
cycles, might be explained by the solar rotation. This picture is also corroborated by the
18.6-day peak in the horizontal component of Reynolds-stress periodogram. This value is
roughly 2/3 of the solar-rotation period.
Although a bit far-fetched, this might seem a more plausible explanation than the Rossby
waves, especially since, at least at first glance, it seems that sub-surface oscillations would
affect sunspot-group positions more than numbers. However, although the motions due to
the Rossby modes with m =  are predominantly toroidal and the modes are long-lived,
the m =  are supposed to cascade to increasingly higher -values where about 50% of the
energy is quickly dissipated (Löptien et al., 2018). Consequently, they are short-lived com-
pared to m =  modes, and this might be the reason why only m =  are observed. If there
indeed are m =  modes on the Sun, then it might be that they modulate the flux emergence
or fragmentation and that their influence can be seen in the sunspot number. Alternatively,
the toroidal motions of m =  modes might affect the solar plasma in such a way that it
results in changes of the flux tubes’ buoyancy leading to increased or decreased sunspots’
appearance depending on the phase of the Rossby mode. In this way the frequencies of
m =  modes might be observed directly in changes of the sunspot number. Finally, simi-
larly to the way that atmospheric waves influence the Earth’s weather, the toroidal motions
of the long-lived m =  modes can influence solar activity and lead to formation of active
longitudes.
To further check the idea that the quasi-annual variability in sunspot number is caused
by Rossby waves, in Figure 6 we plotted the frequencies of the 14 most significant peaks
(FAP < 10%) found in the sunspot-number periodogram with periods between 230 and 400
39 Page 16 of 23 D. Ruždjak et al.

days against the azimuthal wavenumber. The azimuthal wave-number was determined by the
closest matching frequency for the most significant peak at 0.002507 day−1 (399 days), and
then each successive peak attaining less than 10% FAP was considered as the previous mode.
Just for curiosity, the Rieger period was added with m = 11. The line represents the Rossby
waves’ dispersion relation in the solar rest frame. The match is pretty good, although some
points deviate more than 3σ from the theoretical curve; however, the starting mode could
be misidentified and there might be false modes or modes missing in our periodogram. The
FAP, that is used as estimate of the peak significance, throughout this work provides only
the information about the probability that a signal with no periodic component would lead to
a peak of certain magnitude. It does not provide the probability that this data set is periodic
with given frequency and tells us nothing about the rate at which we would expect to miss a
periodic signal that is present in the data.
Consequently, the quasi-annual variability of sunspot number found can be explained as a
consequence of Rossby modes with m = , 17 < m < 30. Rossby waves presumably interact
with convective flows on comparable scales. The turbulence transition to larger-scale Rossby
waves should take place (Löptien et al., 2018) at wavenumber lR ≈ (R /vt )1/2 = 20, where
R is the solar radius, and vt = 4 m s−1 is the rms turbulence velocity. Here the effects
are observed for modes with wavenumbers larger than 17. This might be the consequence
of over-estimated rotation rate or under-estimated rms turbulence velocity in determining
where Rossby waves start to interact with convective flows. Additionally, temporal varia-
tions of the involved quantities might be important and responsible for the discrepancies.
Our results indicate that the Rieger period (m =  = 11) and shorter periods, as well as
QBO (m =  ≥ 50), might not be connected to Rossby modes with  = m. Although they
can be placed near the line on the plot, we cannot provide any reason why some of the
modes would be present and others would not besides the turbulence transition to larger-
scale Rossby waves.
Another question is if the transformation to the Earth’s frame of reference should be
considered, and the mode frequencies shifted by m × ( − E ). Here −1 E = 365.24 day
denotes the period of the Earth’s revolution. If the correction is applied, the original spectrum
is shifted to higher frequencies with all corresponding periods shorter than 15 days, and in
this case the observed mid-term periodicity of sunspot-number changes cannot be connected
to Rossby waves. However, if we assume that the observed variations of the sunspot number
are indeed caused by Rossby waves, then the oscillations of solar plasma in the solar interior
lead somehow to fragmentation of flux tubes or affect the rate at which flux tubes emerge
on solar surface. Whatever the interaction that leads to changes of sunspot number is, it is
taking place in the solar frame of reference, and therefore the changes of sunspot number
should reflect the undisturbed frequencies of solar oscillations.
While quasi-annual periodicity of sunspot-number changes can be explained by Rossby
modes, the changes of the rotation rate and Reynolds stress cannot. There is no match
between the frequencies of Rossby modes and the peaks in the horizontal component of
Reynolds-stress tensor periodogram, nor with the peaks of rotation rate. Although the one-
year period found in solar rotation can be connected to the solar-rotation period, there is
no explanation why only the one-year period is seen as a mid-term variation of the solar
rotation, while the sunspot number displays more periods in that range. The one-year pe-
riod is the period of the Earth’s revolution about the Sun, and it is probable that this does
not represent actual solar variations, but is caused by some observational effect. Beck and
Giles (2005) analysing Dopplergrams from the SOHO/MDI instrument, which provides a
very accurate measurement of meridional flows, found a sinusoidal variation with a period
of one year. They identified it to be due to an error in the determination of the solar-rotation
Periodicities in Solar Rotation and Activity and Connection with Rossby Waves Page 17 of 23 39

axis. The phase and amplitude of the annual variation of meridional flows were used to
determine the error in the Carrington elements. Therefore, a one-year period found in the
rotation-rate residual is probably caused by an error in the rotation-axis determination. This,
however, cannot explain the annual periodicity in the sunspot number, which should not be
affected by the position of the solar-rotation axis, especially since sunspots are located at
low latitudes.
Besides Rossby waves, several other mechanisms have been proposed as a possible
cause of the solar mid-term periodicity (see, e.g., review by Bazilevskaya et al., 2014).
For instance, let us consider the quasi-two-year impulse of shear waves (Pataraya and
Zaqarashvili, 1995), which is caused by velocity and magnetic-field disturbances excited
near the global magnetic-field polarity reversal. This mechanism works for only about two
years around solar minima, which is consistent with the distribution of periods in Figure 5
and could give rise to the observed 2.4-year periodicity. However, mid-term variability of
Reynolds stress, let alone sunspot number, is not so confined to minimum instances, es-
pecially during the last two solar cycles, for its variability to be explained solely by this
mechanism. Another mechanism is the tachocline nonlinear oscillations, where periodically
varying energy exchange takes place between the Rossby waves and differential rotation
(Dikpati et al., 2018). Unfortunately, the period found here is out of range of their calcu-
lations, which were made for periods 2 – 20 months, so no comparison with their results is
possible.
Additionally, a secondary dynamo is also proposed as the mechanism behind solar mid-
term variability. This model is based on the idea of two dynamo sources separated in space.
The first source of the dynamo action, a low-frequency component (22-year cycle), is located
near the bottom of the convection zone, and the second high-frequency component (quasi-
biennial cycle) operates near the top (Benevolenskaya, 1998). Berdyugina and Usoskin
(2003) found the flip-flop cycles, which are defined as the 180-degree shift of the active lon-
gitudes with a period of about 3.5 years, too long for the observed rotation changes. Inspect-
ing their power spectra (Figure 8b in Berdyugina and Usoskin, 2003), besides ≈ 0.3 year−1
peak, the significant peaks at approximately 0.2, 0.4, and 0.5 year−1 (5, 2.5, and 2 years
period) can be also noted. The origin of the flip-flop cycles is not fully understood, but it
could be explained to be the result of two different magnetic-dynamo modes, an axisym-
metric mode and a non-axisymmetric higher-order mode with two cycles in spot patterns
(Berdyugina, Pelt, and Tuominen, 2002). Results from recent numerical simulations con-
firm the possibility of multiple dynamo modes (Käpylä et al., 2016) and two types of cy-
cles (Strugarek et al., 2018). The longer cycle originates at the bottom of the convection
zone, while the shorter cycle is generated in the subsurface layers of the simulation domain.
These cycles are nonlinearly coupled in the turbulent-convective envelopes. Simoniello et al.
(2013) have shown that the secondary cycle in the Sun is an additive contribution to the main
cycle, whose signal strength is rather weak. In the Sun, the major spot activity switches be-
tween the active longitudes in about 1.8 – 1.9 years, and on average it has been observed
to make six switches of the active longitude during the 11-year sunspot cycle (Berdyugina
and Usoskin, 2003). These values have been obtained by analyzing Solar Cycles 18 up to
22, whose lengths were shorter than 11 years. Simoniello et al. (2013) found a significant
periodicity at ≈2 years, a bit higher than the flip-flop findings by Berdyugina and Usoskin
(2003), but Solar Cycle 23 lasted longer than usual (about 12.6 years). This is in agreement
with the indication about long-term behaviour of the Schwabe-cycle length from the pre-
vious section, i.e. shortening of cycle length with rise of solar activity level. Also, then all
periodicities between 1.5 and 4 years (corresponding to individual periods of one longitude’s
dominance) might be expected as the visible manifestation of the same physical mechanism,
which explains the quasi-biennial periodicity observed in sunspot-number data.
39 Page 18 of 23 D. Ruždjak et al.

4. Summary and Conclusion

Solar rotation and activity display variability over a wide range of timescales ranging from
long-term, centennial changes to short-term changes, with periods of several days. There
are similarities and differences between the changes of solar activity (daily total sunspot
number) and solar differential rotation (rotation-rate residual and horizontal component of
the Reynolds-stress tensor).
We found that the long-term changes of all quantities are dominated by the Schwabe
cycle, upon which some changes with longer timescale of unknown character (cyclic or
stochastic) are superimposed. Most probably those are signatures of the Gleissberg cy-
cle. The most notable feature in periodograms is that the activity changes are more com-
plex than rotation changes, and rotation-rate changes are more complex than differential-
rotation changes with more significant periods of changes. The horizontal component of the
Reynolds stress displays the simplest variation pattern.
By analysing the spectral window it was demonstrated that the periods in the range 2.5 to
10 years are not reliable due to influences from aliasing caused by the Schwabe cycle, and
long-term changes and harmonics of quasi-annual and -biennial periods. This does not mean
that there are no variations with those periods, but the period cannot be precisely determined.
Consequently, the algorithms for detecting and characterizing periodic signals in complex
periodic data in astronomy must be used with caution.
It was found that the frequencies of the quasi-annual variations of solar activity with
periods from 230 to 400 days match those of the classical Rossby modes with m ≥ 17. This
is in agreement with the assumption that the turbulence transition to larger-scale Rossby
waves should take place around wavenumber R ≈ 20 (Löptien et al., 2018). The lower
modes are not visible in our data, and variations with shorter periods are most probably
connected to the solar rotation. The match is found only for the unperturbed frequencies
of the Rossby waves. If the frequencies are transformed from solar to the Earth’s frame of
reference, the observed match is ruined.
There is no match between the frequencies of Rossby modes and the peaks in the hor-
izontal component of the Reynolds-stress tensor periodogram. Similarly, the variations of
solar-rotation rate found in our dataset cannot be connected to Rossby waves and are most
probably connected with the existence of active longitudes. Several possible mechanisms
are suggested as possible causes of the mid-term variations in solar rotation.
In our analysis we found strong annual periodicity. This is, in the case of solar rotation,
most probably an observational artifact of the error in the determination of the rotation axis.
To further investigate a connection between quasi-annual periodicity of solar activity and
oscillations in solar interior, we plan to use other datasets, primarily coronal bright-point
data.

Appendix

Here we discuss the effects of the data sampling and processing on the result of the anal-
yses. We consider only the rotation-rate residual and Reynolds-stress data, since those are
the datasets that we created, and the sunspot-number dataset was taken from WDC-SILSO,
ROB. The rotation rates of sunspot groups were calculated from daily shifts by dividing
the differences of CMDs given in the catalogs by elapsed time between successive measure-
ments (usually around one day). Similarly, the meridional motions were calculated by divid-
ing the difference of latitudes by the elapsed time. In this way we obtained synodic-rotation
Periodicities in Solar Rotation and Activity and Connection with Rossby Waves Page 19 of 23 39

Figure 7 Rotation-rate residual and horizontal component of Reynolds stress. Individual data, daily averages,
and monthly averages (from top to bottom), with the corresponding Fourier power spectra.

velocities, which were then transformed to sidereal ones, and the mean rotation curve in the
form of Equation 2 was calculated yielding A = 14.483 ± 0.005 and B = −2.67 ± 0.05.
The rotation-rate residuals (RR) were then calculated by subtracting the mean curve from
individual measurements (ω = ω − (A + B sin2 b)). In this way, the latitudinal dependence
was removed from rotation data. The horizontal component of the Reynolds stress (RS) was
calculated by multiplying the RR with the corresponding values of the meridional motions.
Those data are still latitudinally dependent. Since meridional motions (and therefore RS)
are asymmetric around the Equator, the data were symmetrized by changing the sign for the
southern hemisphere. This is equivalent to changing the sign of meridional velocities for the
southern hemisphere and switching from the North–South direction of the velocity (and RS)
to directions toward the Pole and toward the Equator. The hemispherical dependence on the
latitude still remains in the RS dataset.
The 150,707 individual datapoints are presented in the top panels of the left-hand side of
Figure 7. From those data, 52,100 daily (11 May 1874 to 31 December 2016, including days
with no sunspots) and 1713 monthly averages (May 1874 to December 2016) were calcu-
lated. The averages are given in the next two panels. The image is reminiscent of the butterfly
39 Page 20 of 23 D. Ruždjak et al.

Figure 8 Sequence of wavelet power spectra of horizontal component of Reynolds stress using Morelet
wavelet functions. As before, the solid lines on wavelet spectra denote the COI lines. The horizontal dashed
line denotes the one-year period and vertical ones denote epochs of solar minima. Also, note the different
colour scale than before. In the first panel all data were used. In the second panel the data from minimum years
were omitted. In the third panel the data from minimum years and one year before and after were omitted.
Finally, in the bottom panel the data from minimum years and two years before and after were omitted.
Periodicities in Solar Rotation and Activity and Connection with Rossby Waves Page 21 of 23 39

diagram and the higher scatter after 1976 is due to change between GPR and DPD. By tak-
ing the daily and monthly averages, the picture changes to its “antiopode”. The “noisy” parts
became narrow while the previously narrow parts, which correspond to solar minima, be-
come noisy. This is less dramatic in the case of RR since the signal-to-noise ratio is better. It
is stated in the GPR that the given daily positions of the centers of sunspot groups are sup-
posed to have accuracy of 0.1◦ . However, this estimate is a very optimistic assumption and
the data catalogues are less accurate. Let us, for simplicity, suppose that the data are given
with the accuracy of the order of 1◦ . Now the solar-rotation rate is in the order of 10◦ day−1
and the error of individual values due to errors in position will be on the order of 10%. In
the case of meridional motions which are on the order of 1◦ day−1 the error of individual
values will be around 100%. RS, being the product of RR and meridional motion, will also
have errors of the order of 100%.
Despite the large noise in the data, the dataset is suitable for statistical analyses. For
instance the increase of solar-rotation rate during the minimum of activity can be clearly
seen in the monthly averages of RR, and even noted in daily averages. This is much less
clear in the case of RS, and it is questionable if the observed changes during the solar minima
represent real changes or random fluctuations due to errors. We will return to this question,
but let us examine the effect of averaging on power spectra. By examining the power spectra
on the right-hand side of Figure 7, it can be seen that averaging gives rise to the relative
importance of cycle-related and mid-term changes. This is expected, since by averaging, the
scatter of data is decreased, and the amplitude of real cycle-related variations is becoming
more important. In addition, random noise due to the larger influence of errors caused by
the lack of sunspots in epochs of activity minima could contribute to false signals that are
correlated to the solar cycle. In some cases, some other peaks are popping out, but their
FAP is quite large, except for the 18.6-day period in the case of RS data. Other than this, no
other changes are caused due to averaging. Of course, by averaging, the sampling of data
is changed and the spectral window changes, so to properly analyse the periodicities, the
window should be taken into account.
To examine further if the changes of RS during the solar minima are representing the
real changes or are caused by random errors of the mean due to lack of sunspots during
solar minima, we constructed a sequence of wavelet power spectra (Figure 8) using RS
data. In the first case, all data were used. Next, the data from minimum years were omitted.
In the following step, the data from minimum years and the years before and after were
omitted. Finally, the data from minimum years and two years before and after were omitted,
therefore omitting data from five consecutive years, which represents almost half of the
dataset. Inspecting the sequence, it can be seen that the annual and biennial changes are
connected to the solar minima until the last image. where roughly half of the dataset was
omitted. These results mean that changes of RS are inherent not only in the minimum, but
also in the maximum of solar activity.
To conclude, we will point out that the results are dependent on the data at hand, i.e. the
features of tracers used. Since we are using sunspots, which are less abundant (if not absent)
during solar-activity minima, the results reflect this fact and may be less reliable during solar
minima.

Acknowledgments This work was supported in part by the Croatian Science Foundation under the project
7549 “Millimeter and submillimeter observations of the solar chromosphere with ALMA”. Also, the support
from the Austrian-Croatian Bilateral Scientific Projects “Comparison of ALMA observations with MHD-
simulations of coronal waves interacting with coronal holes” and “Multi-Wavelength Analysis of Solar Rota-
tion Profile” is acknowledged. The support by the Horizon 2020 project SOLARNET (824135, 2019 – 2023)
is acknowledged. We would also like to thank the anonymous reviewers for comments and recommendations
that helped us to improve the readability and quality of the article.
39 Page 22 of 23 D. Ruždjak et al.

Data Availability The datasets generated during and/or analysed during the current study are available from
the corresponding author on reasonable request.

Declarations

Competing Interests The authors declare that they have no conflicts of interest.

References
Balthasar, H., Vazquez, M., Wöhl, H.: 1986, Differential rotation of sunspot groups in the period from 1874
through 1976 and changes of the rotation velocity within the solar cycle. Astron. Astrophys. 155, 87.
ADS.
Baranyi, T., Győri, L., Ludmány, A.: 2016, On-line tools for solar data compiled at the Debrecen observatory
and their extensions with the Greenwich Sunspot Data. Solar Phys. 291, 3081. DOI. ADS.
Bazilevskaya, G., Broomhall, A.-M., Elsworth, Y., Nakariakov, V.M.: 2014, A combined analysis of the
observational aspects of the quasi-biennial oscillation in solar magnetic activity. Space Sci. Rev. 186,
359. DOI. ADS.
Beck, J.G., Giles, P.: 2005, Helioseismic determination of the solar rotation axis. Astrophys. J. Lett. 621,
L153. DOI. ADS.
Benevolenskaya, E.E.: 1998, A model of the double magnetic cycle of the Sun. Astrophys. J. Lett. 509, L49.
DOI. ADS.
Berdyugina, S.V., Pelt, J., Tuominen, I.: 2002, Magnetic activity in the young solar analog LQ Hydrae. I.
Active longitudes and cycles. Astron. Astrophys. 394, 505. DOI. ADS.
Berdyugina, S.V., Usoskin, I.G.: 2003, Active longitudes in sunspot activity: century scale persistence. Astron.
Astrophys. 405, 1121. DOI. ADS.
Brajša, R., Ruždjak, D., Wöhl, H.: 2006, Temporal variations of the solar rotation determined by sunspot
groups. Solar Phys. 237, 365. DOI. ADS.
Brajša, R., Wöhl, H., Vršnak, B., Ruždjak, D., Sudar, D., Roša, D., Hržina, D.: 2002, Differential rotation of
stable recurrent sunspot groups. Solar Phys. 206, 229. DOI. ADS.
Deng, L.H., Zhang, X.J., Deng, H., Mei, Y., Wang, F.: 2020, Systematic regularity of solar coronal rotation
during the time interval 1939 – 2019. Mon. Not. Roy. Astron. Soc. 491, 848. DOI. ADS.
Dickinson, R.E.: 1978, Rossby waves–long-period oscillations of oceans and atmospheres. Annu. Rev. Fluid
Mech. 10, 159. DOI. ADS.
Dikpati, M., McIntosh, S.W., Bothun, G., Cally, P.S., Ghosh, S.S., Gilman, P.A., Umurhan, O.M.: 2018, Role
of interaction between magnetic Rossby waves and tachocline differential rotation in producing solar
seasons. Astrophys. J. 853, 144. DOI. ADS.
Gilman, P.A.: 1968, The general circulation of the solar atmosphere: large thermally driven Rossby waves.
Astron. J. 73, 61. ADS.
Gilman, P.A.: 1969a, A Rossby-wave dynamo for the Sun, I. Solar Phys. 8, 316. DOI. ADS.
Gilman, P.A.: 1969b, A Rossby-wave dynamo for the Sun, II. Solar Phys. 9, 3. DOI. ADS.
Győri, L., Ludmány, A., Baranyi, T.: 2017, Comparative analysis of Debrecen sunspot catalogues. Mon. Not.
Roy. Astron. Soc. 465, 1259. DOI. ADS.
Hathaway, D.H.: 2015, The solar cycle. Liv. Rev. Solar Phys. 12, 4. DOI. ADS.
Holton, J.R.: 1992, An Introduction to Dynamic Meteorology, International Geophysics Series, Academic
Press, San Diego, 216. ADS.
Käpylä, M.J., Käpylä, P.J., Olspert, N., Brandenburg, A., Warnecke, J., Karak, B.B., Pelt, J.: 2016, Multiple
dynamo modes as a mechanism for long-term solar activity variations. Astron. Astrophys. 589, A56.
DOI. ADS.
Khomenko, E., Collados, M.: 2015, Oscillations and waves in sunspots. Liv. Rev. Solar Phys. 12, 6. DOI.
ADS.
Lenz, P., Breger, M.: 2014, Period04: Statistical analysis of large astronomical time series. Astrophys. Source
Code Lib. www.ascl.net/1407.009, ADS.
Liang, Z.-C., Gizon, L., Birch, A.C., Duvall, T.L.: 2019, Time-distance helioseismology of solar Rossby
waves. Astron. Astrophys. 626, A3. DOI. ADS.
Löptien, B., Gizon, L., Birch, A.C., Schou, J., Proxauf, B., Duvall, T.L., Bogart, R.S., Christensen, U.R.:
2018, Global-scale equatorial Rossby waves as an essential component of solar internal dynamics. Nat.
Astron. 2, 568. DOI. ADS.
Periodicities in Solar Rotation and Activity and Connection with Rossby Waves Page 23 of 23 39

McIntosh, S.W., Leamon, R.J., Krista, L.D., Title, A.M., Hudson, H.S., Riley, P., Harder, J.W., Kopp, G.,
Snow, M., Woods, T.N., Kasper, J.C., Stevens, M.L., Ulrich, R.K.: 2015, The solar magnetic activity
band interaction and instabilities that shape quasi-periodic variability. Nat. Commun. 6, 6491. DOI.
ADS.
McIntosh, S.W., Cramer, W.J., Pichardo Marcano, M., Leamon, R.J.: 2017, The detection of Rossby-like
waves on the Sun. Nat. Astron. 1, 0086. DOI. ADS.
Mursula, K., Zieger, B., Vilppola, J.H.: 2003, Mid-term quasi-periodicities in geomagnetic activity during
the last 15 solar cycles: connection to solar dynamo strength to the memory of Karolen I. Paularena
(1957 – 2001). Solar Phys. 212, 201. DOI. ADS.
Olemskoy, S.V., Kitchatinov, L.L.: 2005, On the determination of meridional flow on the Sun by the method
of tracers. Astron. Lett. 31, 706. DOI. ADS.
Pataraya, A.D., Zaqarashvili, T.V.: 1995, Generation of waves by the solar magnetic field polarity reversal.
Solar Phys. 157, 31. DOI. ADS.
Rieutord, M.: 2009, Approaching the low-frequency spectrum of rotating stars. In: Rozelot, J.P., Neiner, C.
(eds.) The Rotation of Sun and Stars Lecture Notes in Physics 765, Springer, Heidelberg, 101. DOI.
ADS.
Rossby, C.G.: 1939, Relation between variations in the intensity of the zonal circulation of the atmosphere
and the displacements of the semi-permanent centers of action. J. Mark. Res. 2, 38.
Roša, D., Brajša, R., Vršnak, B., Wöhl, H.: 1995, The relation between the synodic and sidereal rotation
period of the Sun. Solar Phys. 159, 393. DOI. ADS.
Ruždjak, D., Brajša, R., Sudar, D., Skokić, I., Poljančić Beljan, I.: 2017, A relationship between the solar
rotation and activity analysed by tracing sunspot groups. Solar Phys. 292, 179. DOI. ADS.
Saio, H.: 2018, R mode oscillations ubiquitous in stars. ADS. arXiv.
SILSO World Data Center: 1874 – 2016, The international sunspot number. International Sunspot Number
Monthly Bulletin and online catalogue. www.sidc.be/silso/.
Simoniello, R., Jain, K., Tripathy, S.C., Turck-Chièze, S., Baldner, C., Finsterle, W., Hill, F., Roth, M.: 2013,
The quasi-biennial periodicity as a window on the solar magnetic dynamo configuration. Astrophys. J.
765, 100. DOI. ADS.
Skokić, I., Brajša, R., Roša, D., Hržina, D., Wöhl, H.: 2014, Validity of the relations between the synodic and
sidereal rotation velocities of the Sun. Solar Phys. 289, 1471. DOI. ADS.
Stix, M.: 2002, The Sun: An Introduction (2nd ed.), Springer, Berlin, Heidelberg, 278. ADS.
Strugarek, A., Beaudoin, P., Charbonneau, P., Brun, A.S.: 2018, On the sensitivity of magnetic cycles in
global simulations of solar-like stars. Astrophys. J. 863, 35. DOI. ADS.
Sudar, D., Skokić, I., Ruždjak, D., Brajša, R., Wöhl, H.: 2014, Tracing sunspot groups to determine angular
momentum transfer on the Sun. Mon. Not. Roy. Astron. Soc. 439, 2377. DOI. ADS.
Sudar, D., Brajša, R., Skokić, I., Poljančić Beljan, I., Wöhl, H.: 2017, Meridional motion and Reynolds stress
from Debrecen photoheliographic data. Solar Phys. 292, 86. DOI. ADS.
Usoskin, I.G.: 2017, A history of solar activity over millennia. Liv. Rev. Solar Phys. 14, 3. DOI. ADS.
Usoskin, I.G., Berdyugina, S.V., Poutanen, J.: 2005, Preferred sunspot longitudes: non-axisymmetry and
differential rotation. Astron. Astrophys. 441, 347. DOI. ADS.
Zaqarashvili, T.V., Carbonell, M., Oliver, R., Ballester, J.L.: 2010a, Magnetic Rossby waves in the solar
tachocline and Rieger-type periodicities. Astrophys. J. 709, 749. DOI. ADS.
Zaqarashvili, T.V., Carbonell, M., Oliver, R., Ballester, J.L.: 2010b, Quasi-biennial oscillations in the solar
tachocline caused by magnetic Rossby wave instabilities. Astrophys. J. Lett. 724, L95. DOI. ADS.
Zaqarashvili, T.V., Oliver, R., Hanslmeier, A., Carbonell, M., Ballester, J.L., Gachechiladze, T., Usoskin, I.G.:
2015, Long-term variation in the Sun’s activity caused by magnetic Rossby waves in the tachocline.
Astrophys. J. Lett. 805, L14. DOI. ADS.
Zhang, L., Mursula, K., Usoskin, I., Wang, H.: 2011a, Global analysis of active longitudes of solar X-ray
flares. J. Atmos. Solar-Terr. Phys. 73, 258. DOI. ADS.
Zhang, L., Mursula, K., Usoskin, I., Wang, H.: 2011b, Global analysis of active longitudes of sunspots.
Astron. Astrophys. 529, A23. DOI. ADS.

Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional affiliations.

Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a pub-
lishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript
version of this article is solely governed by the terms of such publishing agreement and applicable law.

You might also like