Relationship between Solar Wind−Magnetosphere Energy and Eurasian Winter Cold Events

Despite global climate warming, boreal winters have experienced frequent cold extremes with heavy snowfall in the early 21st century (Liu et al., 2012; Screen, 2014, 2015; Sun et al., 2016; Francis, 2017). Additionally, a remarkable surface cooling trend appears over Northern Hemisphere midlatitudes after the late-1990s (Cohen et al., 2014; Kug et al., 2015). Many studies have explored possible reasons for such abnormal climate, such as Arctic sea-ice loss and Arctic warming (Honda et al., 2009; Mori et al., 2014; Wang and Liu, 2016; Screen, 2017b; Xu et al., 2018a, b, 2019) and internal atmospheric variability (Screen et al., 2014; McCusker et al., 2016; Screen, 2017a; Ogawa et al., 2018). As the fundamental source of the climate system, solar activity has a profound impact on the decadal variability of boreal winter climate (Gray et al., 2010).

A mounting body of observational and modeling evidence converges on a common conclusion that the occurrence of European cold winters becomes high when the solar activity reaches its minimum on the decadal time scale (Lockwood et al., 2010; Woollings et al., 2010; Ineson et al., 2011; Matthes, 2011). The surface pressure pattern in response to the 11-year solar cycle projects onto the North Atlantic Oscillation (NAO) (Kodera, 2003; Thiéblemont et al., 2015), a dominant mode of the Northern Hemisphere atmospheric variability influencing winter weather and climate in the Eurasia−Atlantic sector (Wallace, 2000; Bader et al., 2011; Matthes, 2011; Cohen et al., 2014; Screen, 2017a). The physical link between the solar ultraviolet (UV) forcing and the decadal NAO variability has been suggested to arise from the “top-down” mechanism (Kodera and Kuroda, 2002; Gray et al., 2010). A lagged and amplified NAO response to the solar cycle is further proposed involving the ocean−atmosphere interaction mechanism (Gray et al., 2013; Scaife et al., 2013). Thiéblemont et al. (2015) revealed a 1−2-year lagged solar−NAO linkage via a multidecadal experiment with 11-year solar forcing variability. Sunspot number, open solar flux, and F10.7 cm radio flux, representing the decadal solar variability or solar UV variations, have been widely used to investigate the solar contribution to Earth’s climate on the decadal time scale (Roy and Haigh, 2010; Woollings et al., 2010; Thiéblemont et al., 2015; Miao et al., 2018). However, the potential connections on the interannual time scale remain unclear.

In addition to radiative forcing, another important aspect of the solar activity influencing Earth’s climate is energetic particle forcing (Lilensten et al., 2015). Matthes et al. (2017) recently suggested that changes in solar spectral irradiance (most importantly in the UV) and energetic particle precipitation (EPP) are two major players for the climate from the aspect of solar variability. The EPP consists of particles originating from the sun, Earth’s magnetospheric field, and from beyond the solar system. Changes in EPP can lead to the production of NO_x through ionization and dissociation and the subsequent intricate chemical processes in the middle atmosphere (Jackman et al., 2006; Rozanov et al., 2012). Then, the polar winter descent of EPP-generated NO_x into the stratosphere affects the ozone abundance (Baumgaertner et al., 2011; Matthes et al., 2017). Sinnhuber et al. (2018) suggested that stratospheric EPP signals are more likely a result of dynamic processes than caused by downward descent of EPP-induced chemistry from the mesosphere and lower thermosphere. Rozanov et al. (2005) provided evidence that EPP-generated variations in stratospheric and tropospheric temperatures are comparable to those caused by solar UV forcing. The possible mechanisms linking the influence of EPP on the stratosphere to the surface are further suggested to be similar to those related to solar UV influence, including the accelerated stratospheric polar night jet and top-down coupling (Seppälä and Clilverd, 2014). Cnossen et al. (2016) also discussed the possible mechanism of downward connection from the thermosphere to the troposphere. However, a modeling study found that the EPP effect on tropospheric temperature is small and insignificant (Meraner and Schmidt, 2018).

As reviewed by Matthes et al. (2017), the influence of different EPP components on the surface climate is an emerging research topic, the challenge of which, however, is to quantify its climate impact. The solar wind, which cannot penetrate into the lower atmosphere directly but can violently interact with Earth’s magnetosphere, serves as a source of EPP (Mironova et al., 2015). The solar wind resembles the high-speed stream generated from the solar coronal regionand mainly encompass electrons, protons, and α-particles. Lu et al. (2008) found a significant relationship between the solar wind and the Northern Annular Mode. The energy input from the solar wind into Earth’s magnetosphere can cause space weather phenomena, such as magnetic storms, aurora, and so on (Akasofu, 1981). Magnetospheric particle precipitation associated with high-speed solar wind streams is more frequent during the decaying phase and near the minimum of the solar cycle (Sinnhuber et al., 2012).

Early studies suggest that the solar wind energy input is largely determined by the solar wind velocity (Crooker et al., 1977) or by the interplanetary magnetic field (IMF) (Dungey, 1961). Different opinions have been proposed since, but it remains a big challenge to quantitatively estimate the solar wind energy input into Earth’s magnetosphere (E_in) (Akasofu, 1981; Newell et al., 2008). Recently, Wang et al. (2014) quantified E_in as a function of interplanetary and solar wind conditions, using three-dimensional magnetohydrodynamics. The new E_in index shows both quasi-decadal variability and interannual variability (He et al., 2018, 2019b), which provides a direct motivation for us to explore the potential interannual connections between the solar activity−controlled magnetospheric energetic particle forcing and the surface climate. Furthermore, He et al. (2018) revealed for the first time that E_in has a tangible effect on the interannual variability of the subsequent winter ENSO. With this in mind, is it plausible that the year-to-year variability of Eurasian winter weather and climate is linked to the solar wind energy penetrating Earth’s magnetosphere. Thus, in this study, we attempt to understand the implications of the preceding E_in for winter extreme cold temperatures over Eurasia on the interannual time scale.

E_in (units: W) is estimated by a three-dimensional magnetohydrodynamic simulation (Wang et al., 2014),

where n_SW (units: cm⁻³) is the solar wind number density; V_SW (units: km s⁻¹) is the solar wind velocity; B_T (units: nT) is the transverse magnetic field magnitude; θ is the IMF clock angle; and the solar wind data can be taken from the NASA OMNI project since 1963 (http://omniweb.gsfc.nasa.gov/). The energy input is more sensitive to the solar wind velocity and the IMF clock angle than other parameters (Wang et al., 2014). The monthly E_in is derived from the daily E_in. The annual mean E_in is employed to show the cumulative effect of energy input from the solar wind into Earth’s magnetosphere. The annual mean E_in index is normalized. In the composite analysis, the higher (lower) E_in years are defined when the normalized annual mean E_in index is equal to or greater (less) than 0.5 (−0.5) standard deviation (Fig. 1c).

Figure 1.  (a) Regressions of frequency (units: d) of winter cold events during 1964−2018 against the normalized annual mean E_in index during 1963−2017. Dotted values are significant at the 95% confidence level based on the Student’s t-test. (b) First mode of the EOF analysis for the frequency of winter cold events over Eurasia (20°−80°N, 0°−150°E) during 1964−2018. Red and blue boxes in (a) and (b) mark the regions where northern Eurasia (NE: 45°−70°N, 40°−140°E) and northern Africa (NA: 20°−45°N, 0°−40°E) are defined, respectively. (c) Normalized annual mean E_in index during 1963−2017 (bars) and normalized frequency of winter cold events during 1964−2018 over NE (blue lines) and NA (red lines). The dots (stars) indicate those years when the annual mean E_in index is equal to or greater (less) than 0.5 (−0.5) standard deviation. The correlation coefficients between the annual mean E_in index and frequency of winter cold events over NE and NA are given at the top of the panel

Daily surface air temperature (SAT) and atmospheric circulation data are derived from the NCEP−NCAR Reanalysis 1 with a horizontal resolution of 2.5° × 2.5° (Kalnay et al., 1996). The daily NAO indices are obtained from the Climate Prediction Center of NOAA (http://www.cpc.ncep.noaa.gov/products/precip/CWlink/pna/nao.shtml). In this paper, we focus on the winters of 1964−2018 considering the time range of the available E_in data. The winter of 1964 (for example) refers to December in 1964 and January and February in 1965. The relationship between E_in activity and winter cold events at lag +1 year (for example, the annual mean E_in in 1963 and cold events in winter 1964) is analyzed. All data are detrended before the regression, composite, and correlation analyses. The significance test is based on the standard two-tailed Student’s t-test. The effective degree of freedom is computed as

where N is the number of samples; r₁ and r₂ are the autocorrelation coefficients of the two variables at one time interval (Bretherton et al., 1999).

Cold events are defined when the daily SAT is below the climatological mean SAT by 1.5 or more standard deviation (Thompson and Wallace, 2001). This study focuses on the total number of cold events in winter, as well as the number of extreme circulation events. Changes in extreme circulation events may not be simply identical to changes in the mean climate and are of greater concern to the public (Screen, 2017a). The extreme sea level pressure (SLP) events, 850-hPa wind (UV850; cyclone) events, and 300-hPa zonal wind (U300) events, are defined as the daily SLP/UV850/U300 above its 90th percentile (Pfahl, 2014). The negative (positive) NAO events are defined as intervals in which the normalized daily NAO index is lesser (greater) than or equal to −1.0 (+1.0) standard deviation for at least three consecutive days (Yao and Luo, 2015). Based on this definition, the frequency of negative (positive) NAO events in a winter is obtained by summing the duration of all negative (positive) NAO events in that winter. The 3−8-day bandpass-filtered 300-hPa transient eddies (v^’2) are calculated according to Murakami (1979).

We use the Eliassen−Palm (EP) flux (F) and its divergence (D_F), which are calculated according to Edmon et al. (1980), to diagnose the wave−mean flow interaction (Andrews et al., 1987):

where ρ is the air density; ${r_0}$ is the radius of Earth; φ is the latitude; u and v are the zonal and meridional wind, respectively; θ is the potential temperature; p is the pressure, θ_p = dθ/dp; primes denote zonal deviations; and overbars denote zonal average. To display the EP flux throughout the stratosphere, the vectors are scaled by $\sqrt {1000/p} $ and 1/ρ (Wang et al., 2009). The vertical component is multiplied by 125 (Castanheira and Graf, 2003). An equatorward (upward) EP flux vector corresponds to a poleward eddy momentum (heat) flux, and the zonal-mean zonal flow decelerates (accelerates) where the EP flux converges (diverges) (Hartmann et al., 2000).

The frequency of Eurasian cold events in the subsequent winter regressed onto the annual mean E_in index is firstly presented in Fig. 1a. Clearly, there is a significantly decreased frequency of cold extremes over northern Asia, Russia and northern Scandinavia, primarily north of 50°N, and an increased frequency over northern Africa and near the Black and Caspian seas (Fig. 1a). The relationship between E_in activity and winter cold events at lag 0 or +2 years is weaker and less significant (Fig. 2). To further investigate the linkage between E_in and the interannual variability of extreme cold weather in the subsequent winter, the dominant interannual modes of cold events over Eurasia (20°−80°N, 0°−150°E) during 1964−2018 are extracted by performing empirical orthogonal function (EOF) analysis. The first mode of EOF (EOF1) explains 23.8% of the total variance, which is 8.6% larger than the second mode. The corresponding principal component of the EOF1 mode (PC1) shows remarkable interdecadal and interannual variability (figure not shown). EOF1 exhibits consistent negative values across the northern Eurasian continent and positive values in northern Africa (Fig. 1b). This pattern of cold events is quite similar to that following higher E_in in the preceding year, with a high spatial correlation coefficient of 0.79. The regions that cover (45°−70°N, 40°−140°E) and (20°−45°N, 0°−50°E) (red and blue rectangles in Figs. 1a and b) are hereafter referred to as northern Eurasia (NE) and northern Africa (NA), respectively. The linear correlation between the annual mean E_in index and the frequency of cold events over NE and NA are −0.38 and 0.30, respectively, significant at the 99% confidence level (Fig. 1c). Moreover, the probability for northern Eurasian winter to experience more than 30 extreme cold days following a higher (lower) E_in year is about 26.7% (48.2%), based on the probability density function (PDF) of the occurrence of cold extremes; the PDF for cold days in NA at lag +1 year shows more occurrences of more than 30 extreme cold days (from 33.0% to 49.4%), in response to higher E_in relative to lower E_in (figure not shown). It is thus hypothesized that the magnetospheric particle precipitation from solar wind is associated with the interannual variations of Eurasian cold events in the subsequent winter.

Figure 2.  Regressions of the frequency (units: d) of winter cold events during (a) 1963−2016 and (b) 1965−2018 against the normalized annual mean E_in index during 1963−2016. Dotted values are significant at the 95% confidence level based on the Student’s t-test

Extreme cold temperatures over Eurasia are affected by anomalous circulations that are related to various forcing (Liu et al., 2012; Cohen et al., 2014). Figure 3a shows a decrease in the frequency of SLP extremes associated with PC1 in the Arctic region and a pronounced increase in the frequency at midlatitudes of the Europe−Atlantic sector (Fig. 3a). More frequent days with the lower-level anomalous anticyclone appear at midlatitudes (Fig. 3b). In the upper troposphere, more (less) frequent extreme strong westerly wind is shown to the north (south) of the climatological westerly jet (Fig. 3c), consistent with the significant positive (negative) U300 anomalies in the north of 50°N (south of 40°N) (figure not shown). This suggests the poleward shift of the North Atlantic westerly jet. The more (less) frequent strong westerly wind near the Norwegian (Mediterranean) Sea would drive milder (colder) conditions in NE (NA). Clearly, the anomalous circulation extremes related to the higher-than-normal E_in in the preceding year (Figs. 3d-f) are qualitatively in good agreement with those related to PC1 (Figs. 3a-c). This provides further support for the interannual connections between the solar wind−magnetosphere energy flux, the subsequent winter circulation extremes in the Europe−Atlantic sector, and extreme temperatures over Eurasia.

Figure 3.  (a−c) Regressions of the frequency (units: d) of winter (a) SLP, (b) UV850, and (c) U300 extremes during 1964−2018 against the normalized PC1 during 1964−2018. (d−f) As in (a−c), but regressed against the normalized annual mean E_in index during 1963−2017. (c, f) Green lines represent the climatology of winter U300 during 1964−2018. Vectors and dotted values are significant at the 95% confidence level based on the Student’s t-test

The North Atlantic synoptic-scale eddy variations play an important role in the highly variable midlatitude weather patterns (Cohen et al., 2014). Figure 4 presents the anomalous storm-track activity response to PC1 and the preceding annual mean E_in, separately. As expected, significantly enhanced storm-track activity emerges primarily in the high-latitude North Atlantic (Fig. 4). Previous studies have suggested that positive storm-track activity anomalies are intimately linked to westerly wind anomalies in situ, cyclonic eddy forcing to the north, and anticyclonic eddy forcing to the south (Lau, 1988; Gong et al., 2011; He et al., 2019a). The northward shift of the synoptic-scale eddies favors the northward shift of the North Atlantic westerly jet and more frequent strong westerly wind at high latitudes (Figs. 3c and f). The increases in the strength of the synoptic-scale eddy forcing at high latitudes are also tied to the cyclone extremes to the north and anticyclone extremes to the south (Figs. 3b and e). Considering the wider literature, changes in the North Atlantic storm tracks are consistent with the shift of the NAO phase (Bader et al., 2011; Cohen et al., 2014). A poleward shift of the storm track occurs when the NAO is in its positive phase (NAO|+) and an equatorward shift is observed in negative NAO (NAO|−) winters. The influence of geomagnetic activity on the NAO has been examined in previous studies (Baumgaertner et al., 2011; Li et al., 2011). A dynamical change of the positive shift of the NAO is thus expected in association with the higher solar wind−magnetosphere energy flux input.

Figure 4.  Regressions of winter 3−8-day bandpass-filtered 300-hPa transient eddies (v^′2; units: 10⁻³ m² s⁻²) against (a) the normalized PC1 during 1964−2018 and (b) the normalized annual mean E_in index during 1963−2017. Green lines represent the climatology of winter U300 during 1964−2018. Dotted values are significant at the 95% confidence level based on the Student’s t-test

Next, we turn our attention to the frequency of strong NAO events following higher and lower E_in years. As shown in Fig. 5a, the composite occurrence of 1-year-lagged NAO|+ events is higher-than-normal following higher E_in, and lower-than-normal following lower E_in. Similarly, the NAO|− events have a higher frequency following lower E_in compared to higher E_in (Fig. 5a). These differences support the notion that the subsequent winter NAO|+ (NAO|−) events are intensified (weakened) following higher E_in activity, favoring less frequent cold events in NE and more frequent cold events in NA, and vice versa (Fig. 5b).

Figure 5.  Composites of the frequency of winter (a) NAO|+ and NAO|− events and (b) cold events over northern Eurasia (NE) and northern Africa (NA) during 1964−2018 following the higher (red bars) and lower (blue bars) E_in years during 1963−2017

The potential influence of magnetospheric energetic particle forcing on stratospheric circulation has been suggested in previous studies (Seppälä et al., 2009; Baumgaertner et al., 2011; Rozanov et al., 2012). Associated with the higher energy input from the solar wind into Earth’s magnetosphere, the tropical region warms in the upper troposphere and lower stratosphere in the following year from spring (March−May) to winter (Fig. 6a). This might be due to the cumulative effect of energy input from the solar wind into Earth’s magnetosphere and increased ozone heating (Kodera and Kuroda, 2002; Cionni et al., 2011). Additionally, it is also possible that the relationship between E_in and equatorial temperature is modulated by ENSO, because a previous study revealed a close relationship between boreal winter ENSO and the preceding annual-mean solar wind−magnetosphere energy flux input (He et al., 2018). The westerly wind is profoundly accelerated throughout the troposphere and stratosphere in boreal winter (Fig. 6b), which might be attributable to the intensified northward temperature gradient due to the persistent warming at lower latitudes (Fig. 6a). There are anomalous downward-pointing EP flux vectors from the stratosphere to the upper troposphere in the following December and January (Figs. 7a and b; vectors), suggesting reduced upward Rossby wave propagation from the troposphere into the stratosphere. Also, the EP flux divergent anomalies in the troposphere (Figs. 7a and b; contours) are consistent with westerly anomalies at 20°−35°N and 50°−75°N and easterly anomalies in between (Fig. 3f). This wave−mean flow interaction (Charney and Drazin, 1961; Hartmann et al., 2000) is favorable for maintaining the downward propagation of the solar signals (Thiéblemont et al., 2015) and the connection of E_in signals to tropospheric circulation and climate. It might be the case that there is no clear wave−mean flow interaction associated with E_in in February (Fig. 7c). However, there are some significant SLP anomalies in February (Fig. 8c), consistent with those in December and January (Figs. 8a and b). This might be because the impacts of E_in have propagated into the troposphere in December and January (Figs. 7a and b).

Figure 6.  Vertical−horizontal cross section for the correlations between the daily zonal-mean (a) temperature averaged along 30°S−30°N and (b) zonal wind averaged along 55°−65°N (from 1 March to 28 February, with 5-day low-pass filtering) during 1964/65−2018/19 and the normalized annual mean E_in index during 1963−2017. Dotted values are significant at the 95% confidence level based on the Student’s t-test. The effective degrees of freedom are adopted in (a) for the Student’s t-test

Figure 7.  Composites of the subsequent (a) December, (b) January, and (c) February Eliassen−Palm flux (vectors; units: 10⁷ m² s⁻²) and its divergence anomalies (contours; units: m s⁻¹ d⁻¹; red/blue contours indicate anomalous divergence/convergence) between the higher and lower E_in years during 1963−2017

Figure 8.  Regressions of the subsequent (a) December, (b) January, and (c) February frequency (units: d) of SLP extremes during 1964/65−2018/19 against the normalized annual mean E_in index during 1963−2017. Dotted values are significant at the 95% confidence level based on the Student’s t-test

Solar activity is a major energy source of Earth’s climate, through changes in radiative forcing and energetic particle forcing (Lilensten et al., 2015). The impacts of the variation in solar irradiance on the decadal variability of Eurasian winter climate have been well recorded in the literature (Gray et al., 2010; Woollings et al., 2010; Thiéblemont et al., 2015). However, it remains a big challenge to quantify the influence of the different EPP components on the surface climate, which is an emerging research topic (Matthes et al., 2017).

The solar wind serves as a source of magnetospheric energetic particle forcing, in the main form of electrons (Mironova et al., 2015). Recently, Wang et al. (2014) quantified the solar wind energy input into Earth’s magnetosphere (E_in) using three-dimensional magnetohydrodynamics. It is worth noting that this new E_in index exhibits both interdecadal and interannual variability (He et al., 2018, 2019b). The interannual relationship between the solar wind energy penetrating Earth’s magnetosphere and Eurasian cold extremes in the subsequent winter is thus explored in this study. The results show that the frequency of cold winter events following years with increased solar wind−magnetosphere energy flux shows a significant decrease over NE and increase in NA, and vice versa. This is broadly similar to the dominant EOF mode of interannual variability in Eurasian cold extremes, with a high spatial correlation coefficient of 0.79. The high variance of winter cold events over NE and NA explained by the preceding annual mean E_in can reach up to 0.2 (figure not shown). Moreover, the frequent positive NAO (NAO|+) events and poleward shift of the North Atlantic storm track activities following the preceding higher-than-normal E_in activity, favors the anomalous circulation extremes and less frequent extreme cold temperatures over northern Eurasia. It is further hypothesized that the wave−mean flow interaction in the stratosphere and troposphere is favorable for the connection of E_in signals to tropospheric circulation and climate in the following winter. In general, our study provides further insight into the potential interannual linkage between EPP variations and the surface extreme cold weather.

Acknowledgements. This work was supported by the National Key R&D Program of China (Grant No. 2016YFA0600703), the National Natural Science Foundation of China (Grant Nos. 41875118, 41605059 and 41505073), the Postgraduate Research & Practice Innovation Program of Jiangsu Province (Grant No. KYCX18_0997), the funding of the Jiangsu Innovation & Entrepreneurship Team, and the Priority Academic Program Development (PAPD) of Jiangsu Higher Education Institutions. This work was partially supported by the International Space Science Institute (ISSI) in Beijing, through the working team “Dynamical signatures of energetic particle precipitation in atmospheric re-analyses” (ID:32, 2019).