In this study, we used the monthly land surface precipitation data from the University of Delaware (v5.01), with a horizontal resolution of 0.5°. For the circulation variables, we used the monthly National Centers for Environmental Prediction/National Center for Atmospheric Research (NCEP/NCAR) reanalysis (Kalnay et al., 1996), with a horizontal resolution of 2.5°. The monthly Hadley Centre Sea-ice and Sea-surface Temperature Data Set Version 1 (HadISST1) at 1° resolution are also used in the present study (Rayner et al., 2003). Monthly sunspot numbers (SSN) data set is employed to quantify the solar activity, which is available at http://www.esrl.noaa.gov/psd/gcos_wgsp/Timeseries/SUNSPOT/. The SSN can well represent the variability of total and ultraviolet solar irradiance, both of which are important solar parameters that influence global and regional climate (Gray et al., 2010). In order to characterize the impacts from the major volcanic eruptions, the Northern Hemisphere averaged stratospheric aerosol optical depth (AOD) is used, which can be downloaded from https://data.giss.nasa.gov/modelforce/strataer/. The analysis period of this study is 1948–2016.

Both the solar activity and the climate variables exhibit inter-decadal variability or long-term trends. As seen in Figure 1A, the SSN shows slightly inter-decadal change and decreasing trend during the past decades. But the primary focus of the present study is the 11-years solar cycle modulation on the interannual relationship between ENSO and circulation/precipitation. For this purpose, in this study, the long-term trend, as well as the inter-decadal (greater than 15 years) component of the SSN/precipitation/SST/circulation had been removed. First, all variables are detrended. Then, they are smoothed using Lanczos low-pass filtering so that only time variations greater than 15 years were retained (Duchon 1979). Note that, a 15 years low-pass filtering requires additional 7 years at the beginning and the end of the time series. For variables that are not available before 1948 or after 2016, we extended the time series by repeating the first or last 7 years of the analysis period. This long-term signal is then subtracted from the original time series to obtain a shorter-period component that we refer to as the interannual-decadal component. After these processes, the potential impacts from the inter-decadal variations on composite analysis can be well eliminated, as in shown in Figures 1B–D for the time series of SSN, ENSO and SCP, these adjusted variables mainly show interannual or decadal variations.

In order to investigate the combined effects of the ENSO and the solar activity depending on their different phases, each year has been categorized into four groups (Table 1), i.e., the wENSO/HS years, the wENSO/LS years, the cENSO/HS years and the cENSO/LS years, based on the classification in Section 2. Figures 4A–D show the composites of rainfall anomalies under these four different combinations of ENSO and solar phases. For the wENSO, the rainfall anomalies show notable differences between the HS and the LS phase. Positive January rainfall anomalies appear over most areas of the SC in wENSO/HS years (Figure 4A), with the amplitude exceeding 30 mm/ month around the coastal area of SC. The statistically significant positive signal (90% confidence level) mainly distributes in the provinces of Hunan and Fujian (approximately located at 24°–28°N, 110°–118°E). Thus the rainfall pattern in wENSO/HS years is qualitatively consistent with that in the common wENSO years (as indicated in Figure 2A), featuring a wetter than normal condition over the SC. However, during the wENSO/LS (Figure 4C), the expected wENSO associated wet anomalies are almost missing. Instead, dry anomalies are detected over most of the SC, although with limited statistical significance.

Similarly, the cENSO associated precipitation anomalies are also very different under different solar phases. Generally, cENSO events correspond to dry anomalies over the SC. However, in cENSO/HS years, positive rainfall anomalies dominate the SC, with statistically significant (exceeding 95% confidence level) signals appear in most areas of the YRV region. While in the cENSO/LS years, the cENSO-related dry anomalies become evident. The rainfall pattern features statistically significant negative anomalies in most areas to the south of the Yangtze River. The negative center exceeds −25 mm/ month, locating at the coastal region of the SC. Therefore, we may conclude that when the ENSO and the solar activity are in phase, i.e., the wENSO/HS years and the cENSO/LS years, the ENSO-related precipitation anomalies are intensified. If they are out of phase, the known ENSO-SCP relationship is disturbed.

To further demonstrate the solar impacts on the ENSO precipitation, Figure 4E shows the rainfall difference between the wENSO/HS years and the wENSO/LS years, while the rainfall difference between the cENSO/HS years and the cENSO/LS years is shown in Figure 4F, the solar modulations on both the wENSO and the cENSO precipitation anomalies can be hence inferred. It is found that there are generally consistent patterns of solar influences on the wENSO and the cENSO associated precipitation anomalies, as both Figures 4E,F show mono-pole pattern of positive rainfall differences over the SC. Notably, as had been indicated in Figure 2B, the solar signal in January precipitation exhibits significant positive anomalies over eastern and southern part of China, the pattern of which bears some resemblance with those in Figures 4E,F. Therefore, to some extent, the solar modulations on wENSO and cENSO rainfall in mid-winter are possibly due to the background of solar signal superimposed onto the ENSO signal.

But we also note that the strength of solar modulation effect seems to be different between the wENSO and the cENSO phase. Seen in Figure 4E, although solar activity tends to increase the wENSO precipitation anomalies, the wENSO/HS minus wENSO/LS rainfall anomalies are mostly statistically insignificant. In contrast, the solar effects are stronger in amplitude and are more statistically significant during cENSO phase, with significant rainfall differences between the cENSO/HS years and the cENSO/LS years detected over most of the SC. Therefore, the solar modulating effects show clear asymmetries in strength between the wENSO and the cENSO events.

To further investigate this, we now focus attention on the area-averaged (110–120°E, 22–35°N) SCP anomalies, chosen because it shows evident precipitation difference between HS and LS under both ENSO phases. Scatter plots of standardized January SCP against DJ Niño-3.4 index are shown in Figure 5A, with the HS and LS years expressed by red and blue markers, respectively. Seen from Figure 5A, the correlation coefficient between the Niño 3.4 index and the SCP is 0.31 (p < 0.05). But the SCP anomalies under both wENSO and cENSO phases still show large spread, indicating other factors (e.g. atmospheric internal variability or external forcings) may also impact the precipitation. For wENSO phase, the average SCP is only slightly higher during HS than during LS, suggesting insignificant solar impact under this case. While for the cENSO phase, the HS and LS precipitation anomalies are quite distinguishable, as most cENSO/HS years have positive SCP anomalies while most cENSO/LS years have negative values. Thus solar forcing seems to play an important role under cENSO condition. These results are quite consistent with those derived from the composite analysis, also suggesting solar activity exerts different influences on wENSO and cENSO events. To further demonstrate this, Figures 5B,C show scatter plots of SCP against SSN under wENSO and cENSO years, respectively. The correlation between SCP and SSN is 0.11 and 0.59 (p < 0.01) for the two ENSO phases, respectively. Thus the solar impact on SCP is much stronger under cENSO than in wENSO, which confirms the results of Figures 4E,F.

How can we understand these combined solar/ENSO effects on precipitation? To investigate this, Figures 6A–F depict the composite SLP anomalies over the Northern Hemisphere in different ENSO/solar phase combinations, while Figures 7A–F are the same as Figures 6A–F, but displaying GPH anomalies at 500 hPa. At mid-high latitudes, both wENSO/HS and wENSO/LS years show a pattern that resembles the negative phase of the AO, possibly due to the tropical SST forcing on the polar vortex (Bell et al., 2009). Over the North Pacific, significant low pressure anomalies are seen at both surface and 500 hPa, reflecting the wENSO impacts on the Aleutian low and the PNA pattern, respectively. However, over East Asia, very few statistical significant circulation signals can be detected, except for the wENSO/LS case at 500 hPa. The wENSO/HS minus wENSO/LS circulation over this region exhibits a barotropical structure (consistent pattern at surface and 500 hPa), with positive pressure anomaly over northern East Asia and negative one over southern East Asia, but these anomalies are also with little significance. Considering the limited sample sizes of the wENSO/HS and wENSO/LS years may introduce uncertainties to the composite results, in Figures 6G, 7G the regression of circulation anomalies onto the solar index (i.e., SSN) under wENSO are shown as supplements to the composite analysis. The regression patterns, generally similar to the composite differences, show weak and insignificant solar influences on atmospheric circulation under wENSO phase. Therefore, it’s reasonable to see a relatively weak and statistically insignificant solar impact on SCP under wENSO.

By contrast, the solar modulations on cENSO associated circulation patterns seem to be more clear and evident. In cENSO/HS years, significant negative SLP anomalies are seen in the tropical western Pacific and the Indian Ocean, reflecting a strengthened Walker Circulation associated with cENSO SST anomalies. In the extra-tropics, high pressure anomaly exists over the Aleutian region while low pressure anomaly occupies the Asian continent. Although these pressure systems show limited significance in most regions possibly due to the small sample sizes, we noted a statistically significant low GPH anomaly extending from the Tibet Plateau to Indochina Peninsula at 500 hPa, indicating the strengthening of the southern branch trough. This circulation system had been shown to closely related to the wintertime SCP (Li et al., 2017), which is able to explain the significant positive SCP under cENSO/HS.

The Asia-Pacific circulation patterns are almost reversed in the cENSO/LS years. Positive SLP anomaly is seen over the Asian continent while negative one located in the Aleutian region. Although these circulation anomalies at the surface are with little significance, significant negative GPH anomalies are detected over the East Asian coast at 500 hPa, indicative of an intensified East Asian trough. A deepened East Asian trough usually corresponds to a stronger cold air outbreaks over China, thus can provide reasonable explanation for the significant negative SCP anomalies shown in Figure 4D. As indicated by Figures 6F, 7F, the solar effect on circulation during cENSO resembles the positive phase of AO, similar with the MLR derived solar pattern shown in Figures 3B,D. The regressions of circulation onto SSN under cENSO are shown in Figures 6H, 7H. The derived patterns are similar with the composite difference patterns, showing a tendency towards positive AO under HS, but with much more statistical significance possibly because the sample size is larger. At the surface, significant positive SLP signals exist in both the Aleutian region and the Azores region. In addition, the SSN-AO correlation reaches 0.56 (p < 0.01) under cENSO phase, whereas it drops to near zero under wENSO phase.

Albeit the solar signal under cENSO phase resembles the AO pattern, it also exhibits evident zonal asymmetric component. Over the Asia-Pacific region, the GPH at 500 hPa is significantly increased over the East Asian coast and the North Pacific while significantly decreased over the Asian continent. In order to identify the circulation pattern favorable for the SCP, Figure 8A shows composite of 500 hPa GPH between high and low SCP years. It is found that enhanced SCP is associated with negative GPH anomaly over central-western part of Asia while positive anomaly around Japan, appearing as a “West Low East High (WLEH)” pattern. The WLEH pattern corresponds to a weakened East Asian trough, as well as an intensified trough over the West Asia, which would enhance the southerly along the East Asian coast and thus favoring the SCP. The solar pattern in Figure 7H also resembles WLEH pattern over the Asia-Pacific region, suggesting solar may modulate SCP through exerting a WLEH-like circulation pattern during cENSO. To further demonstrate this, we define a WLEH index (WLEHI) as 500 hPa GPH difference between 120–150°E, 30–40°N and 30–90°E, 22–35°N. The WLEHI is significantly correlated with the SCP (r = 0.55, p < 0.01) when all years in the analysis period are used, demonstrating its important role in the SCP variability. The correlation between SSN and WLEHI is 0.15 (p > 0.1) under wENSO and 0.54 (p < 0.01) under cENSO, thus the WLEH pattern seems to bridge the solar activity and the SCP during cENSO.

Previous studies had noted solar influences in Northern Hemisphere winter circulation. Many of them pointed to a surface response over the extra-tropical regions that bears some resemblance to the AO/NAO pattern (Kodera and Kuroda, 2002; Woollings et al., 2010; Chen et al., 2015). However, the statistical significant signals are mostly found over the Aleutian low region rather than the Arctic region (e.g. Hood et al., 2013). These features revealed in the previous literature can be well reproduced in Figures 3B,D of the present study. The stratosphere provides a key link for solar irradiance variations to interact with the AO/NAO like circulation pattern. In HS years, the enhanced solar UV irradiance can intensify stratospheric zonal winds by affecting the amount of ozone and radiative heating in the mesosphere and the stratosphere. Then the solar-related stratospheric zonal wind anomalies would propagate downward to the surface through wave-mean flow interaction, ultimately impact the phase of the AO/NAO. The above mentioned process is known as the “top-down” mechanism. It is well established and had been demonstrated by a number of modelling studies (Haigh, 1994; Ineson et al., 2011). However, the solar signals are usually not quite strong in amplitude, with an order of approximately 3-4 hPa in each node of the AO/NAO (Gray et al., 2016). Due to the presence of strong internal noises, the derived solar signal is non-stationary and statistically insignificant over the Arctic regions. Instead, robust and significant signal appears at the mid-latitudes, particularly over the North Pacific region. In addition, the feedback from the Pacific and North Atlantic SST anomalies may also act to reinforce and reshape the solar pattern, giving rise to some zonal asymmetric features, including the North Pacific high pressure anomaly and the EATL/WRUS pattern (Brugnara et al., 2013; Hood et al., 2013; Chen et al., 2015).

As revealed in In Sect. 3.3, solar signal in NH circulation depends on the phase of ENSO. The circulation responses to solar activity are relatively weaker and insignificant under wENSO while stronger and significant under cENSO. These asymmetric solar signals in circulation are crucial for the different solar effects on SCP under different ENSO phases. To date, the underlying physical mechanisms remains unclear. Notably, the findings in a recent study of Guttu et al. (2021) provided insight on this issue. They found the “top-down” solar forcing be enhanced (suppressed) under negative (positive) phase of the PDO. Through modulating planetary wave activities, the SST anomalies in negative PDO phase favor negative meridional heatflux at upper troposphere and lower stratosphere, which amplifies the Arctic cooling and intensifies the polar vortex, and hence strengthen the solar-induced AO-like circulation pattern. It is known that the ENSO and PDO SST patterns are very similar, the cENSO can also modulate the upward planetary wave activities and strengthen the polar vortex (Bell et al., 2009). Therefore, the ENSO state may also modulate the “top-down” solar forcing through changing the background state as the PDO does.

Considering the ENSO could influence the East Asian climate by a teleconnection between tropics and extra-tropics (e.g., Liu and Zhu, 2019), thus an “oceanic pathway” mechanism of the solar effect is also possible, in which the distinct ENSO-related SST patterns between HS and LS hold a key. To verify this, the composites of the ENSO associated SST anomalies under different solar phases are displayed in Figure 8. In both HS and LS, the typical ENSO features over the tropical Pacific are evident, with positive (negative) SST anomalies in equatorial central-eastern Pacific and negative (positive) SSTAs in WNP during wENSO (cENSO) events (Figures 9A–D). But differences in SST distributions between the HS and LS years can be noted (Figures 9E,F). Both the SST differences between wENSO/HS years and the wENSO/LS years, and the SST differences between cENSO/HS years and the cENSO/LS years exhibit central-eastern Pacific cooling. In response to such SST anomalies, the zonal gradient of anomalous SLP in the tropical Indo-Pacific Ocean is stronger in the wENSO/LS group than the wENSO/HS one (Figure 6E), whereas the tropical circulation anomalies in the cENSO/HS category are more evident than the cENSO/LS one (Figure 6F). However, we found these solar modulations on ENSO SSTs over tropics unable to explain the observed SCP anomalies. Take the solar influences on cENSO as an example, HS tends to intensified the cENSO associated eastern Pacific cooling and the Walker Circulation, although not quite significantly. As is known, in winter, eastern Pacific cooling and intensified Walker Circulation usually corresponds to an enhanced EAWM and decreased SCP. However, significant increased SCP is seen in cENSO/HS years (Figure 4B).

On the other hand, the SSTs over the mid-latitudes of the North Pacific Ocean may play a role in atmospheric circulation anomalies, since this region is situated at the confluence of warm and cold currents near 40°N, A hot spot region of climate variability. The common wintertime atmospheric responses to the mid-latitude Pacific SST anomalies tend to be an equivalent barotropic structure in the presence of transient eddy forcing, i.e., an anomalous atmospheric ridge over the negative (positive) mid-latitude SST anomalies (e.g., Fang and Yang, 2016). As shown in Figure 9F, the most significant SST differences between HS and LS years emerge in the mid-latitudes of the North Pacific under cENSO condition, which would exert a positive feedback onto the positive SLP/GPH anomalies over the North Pacific. Therefore, it is also possible that solar may modulate the cENSO associated climate effects through mid-latitude air-sea interactions.