Robust Solar Signature in Late Winter Precipitation Over Southern China

The 11‐year solar cycle (SC) has been widely recognized as a potential source of regional climate variability in the Northern Hemisphere winter. However, whether an SC signal exists in the southern China winter precipitation remains unclear. By analyzing land surface precipitation and sunspot number data from 1901 to 2010 in this study, evidence of a robust positive precipitation response that is synchronous with the SC in late winter (January–March) over southern China is provided, and the most statistically significant signals (p < 0.01) are detected over the middle Yangtze River basin. In early winter (October–December), there are only nonsignificant negative responses. The late winter SC‐precipitation relationship persists over time and appears to largely result from a solar‐associated Rossby wave train originating from the North Atlantic/European region. Our results indicate an enhanced predictability of late winter precipitation over southern China.


Introduction
Dominated by the East Asian Winter Monsoon, the most prominent winter characteristic over China is a cold and dry climate. However, severe rainfall/snowstorm events occasionally occur. For example, in January 2008, an intense and long-lasting snowfall/icy rain event occurred over southern China, producing catastrophic consequences for society and the economy (e.g., Wen et al., 2009). Recently, the causes of winter precipitation anomalies over southern China have attracted increasing attention (Jia & Ge, 2017;Zhang et al., 2015;Zhou & Wu, 2010) but have not been as intensively studied as the summer counterpart.
Wintertime precipitation over southern China is known to be strongly affected by tropical sea surface temperature (SST) anomalies. Numerous studies have noted that the El Niño-Southern Oscillation (ENSO; Chen et al., 2014;Wang et al., 2000;Wu et al., 2003;Zhou, 2011), as well as positive SST anomalies in the tropical Indian Ocean Li & Zhou, 2016;Wu et al., 2018), can increase southern China winter precipitation. Apart from tropical SSTs, extratropical atmospheric dynamical processes, particularly teleconnections originating from the North Atlantic/European (NAE) region, also affect precipitation anomalies in China. As revealed by Liu et al. (2014), the circulation anomalies over the NAE region are correlated with China precipitation through three different Eurasian teleconnection patterns. The NAE circulation perturbations can also impact downstream precipitation over China through a subtropical ©2019. The Authors. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.

10.1029/2019GL084083
Key Points: • There is a robust and significant positive correlation between the 11-year solar cycle and southern China precipitation during late winter (January-March) • This late winter SC-precipitation relationship largely results from a solar-associated Rossby wave train originating from the North Atlantic/European region Supporting Information: • Supporting Information S1 • Figure S1 • Figure S2 • Figure S3 • Figure S4 Correspondence pathway (Ding & Li, 2017), which is characterized by a wave train that propagates along the winter Asian jet stream Watanabe, 2004).
Notably, these midlatitude circulation anomalies that influence China precipitation are likely generated by an external forcing to some extent. A growing body of evidence has demonstrated that the 11-year solar cycle (SC) plays an important role in the decadal winter variability over the NAE region (Brugnara et al., 2013;Gray et al., 2010;Kodera et al., 2016;Thiéblemont et al., 2015;Woollings et al., 2010). The SC influences surface climate through various mechanisms. One mechanism is the "bottom-up" mechanism, which refers to amplification of the direct total solar irradiance effect at the ocean surface via moisture transport feedback in the tropical Pacific (Meehl et al., 2008;Meehl & Arblaster, 2009). Another mechanism known as the "topdown" mechanism is triggered by variabilities in solar ultraviolet irradiance, which produce upper stratospheric temperature and circulation anomalies, which are then transferred downward to the surface via wave-mean flow interactions (Chen et al., 2015;Haigh, 1994;Ineson et al., 2011;Matthes et al., 2006). Many recent studies found that the top-down solar forcing signals can be modulated by ocean-atmosphere coupling (e.g., Roy, 2014). The extended memory of the ocean heat-content anomalies and their feedbacks onto the atmosphere induce a lagged solar influence on the Azores mean sea level pressure (Andrews et al., 2015;Gray et al., 2013;Scaife et al., 2013). In the NAE region, the direct top-down solar forcing and solar-induced ocean-atmosphere coupling are dominant during different winter phases, thus producing very different surface climate impacts between early and late winter (Gray et al., 2016;Ma et al., 2018).
Considering the significant SC signals in the NAE region and that China precipitation is teleconnected with circulation anomalies over this region, an important scientific question as to whether an SC signal appears in China winter precipitation is raised. Therefore, the purpose of this study is to determine the SC role in influencing the winter precipitation anomalies over China.

Data
The primary monthly land surface precipitation data used in this study are the Climatic Research Unit (CRU) TS v4.01 (Harris et al., 2014), with a 0.5°resolution from 1901 to 2010. For comparison, we also examined our results with the Global Precipitation Climatology Centre Full Data Product (GPCC) V7 (Schneider et al., 2014) and the precipitation data provided by the University of Delaware (UDEL; Willmott & Matsuura, 2001). All three data sets have a horizontal resolution of 0.5°. Notably, these data show very similar precipitation responses over southern China (see Figure S1 in the supporting information). The homogenized observed monthly precipitation for China from 1951 to 2012 was acquired from the China Meteorological Administration (Yang & Li, 2014). In this study, we also employ the European Centre for Medium-Range Weather Forecasts (ECMWF) 20th century reanalysis (ERA-20C) data set with a horizontal resolution of 2.5°for the 1901-2010 period (Poli et al., 2016). Monthly sunspot numbers (SSNs) are used to quantify the SC (http://www.esrl.noaa.gov/psd/gcos_wgsp/Timeseries/SUNSPOT/). The stratospheric aerosol optical depth (AOD) averaged over the Northern Hemisphere is employed to represent the volcanic influences (downloaded from https://data.giss.nasa.gov/modelforce/strataer/). We used the Nino 3.4 index to represent the ENSO, which is derived from the Hadley Centre Sea Ice and Sea Surface Temperature data set (Rayner et al., 2003).

Methodology
In this study, the SC signal is derived using a multiple linear regression (MLR) analysis. The MLR method has been frequently used to isolate solar influences from other variability sources (Frame & Gray, 2010;Lean & Rind, 2008;Ma et al., 2018;Roy & Haigh, 2010). In the MLR analysis, any climate variable T is assumed to be a function of a space vector x and time and can be expressed as follows: C denotes the regression coefficient of each climate index. The four climate indices introduced into the MLR equation are as follows: (1) SSN, the extended winter (October-March) mean SSN that represents solar activity; (2) AOD, the extended winter mean stratospheric AOD averaged over the Northern Hemisphere that represents volcanic influences; (3) ENSO, the Nino 3.4 index that quantifies ENSO; and (4) TREND, a linear trend term that approximately represents the anthropogenic forcing. The parameter ε represents the residual term of the MLR equation. To address the potential autocorrelation in the residual term, following Chen et al. (2015), a prewhitening procedure is employed three times to ensure that most grids satisfied the Durbin-Watson test. Subsequently, the statistical significance level of the regression coefficients is determined by a 1,000-trial bootstrap resampling test. The estimated SC signals are denoted by the SSN regression coefficients, which were scaled to obtain an estimate of the maximum likely atmospheric responses to the SC.

Results
Recent studies have suggested that the SC impacts on wintertime surface climate have evident subseasonal variations (e.g., Gray et al., 2016). Therefore, to investigate whether there is a subseasonal evolution of the SC-precipitation relationship, Figure 1a shows the time-latitude cross section of SC signals in precipitation averaged between 110°E and 118°E. During the early winter months (October-December, OND), negative precipitation responses can be observed in most of southern China. Notably, the response reverses sign as the season progresses. Positive responses become dominant during the late winter months (January-March, JFM), and the statistically significant signals are mainly detected in January. The spatial distributions of the SC signals in precipitation for the early (OND) and late (JFM) winter averages are further shown in Figures 1b and 1c. Notably, to show the spatial distribution of the precipitation response over the entire central-East China (CEC) region, the latitude range in Figures 1b and 1c is set to 15-60°N. However, we only show the longitudinal averaged precipitation response over 22-40°N in Figure 1a because the area north of 40°N at 110-118°E is beyond the boundary of China. Figure 1b suggests notably dry conditions over southern China during early winter, but in most areas, the signals are statistically nonsignificant. In contrast, a significant SC influence emerges in late winter (Figure 1c). The dominant signal exhibits a region of enhanced precipitation centered over the middle Yangtze River (MYR) basin and extends across most of southern China. The peak response exceeds 20 mm/month and is statistically significant at the 99% confidence level. Such significant precipitation responses in the MYR basin exist not only in CRU data sets but  Figure 1 are millimeters per month. Solid black (white) dots denote regions where the SC signals are statistically significant at the 5% (1%) level (i. e., p < 0.05 (p < 0.01)) after prewhitening and a 1,000-trial bootstrap resampling test. also in the GPCC and UDEL data sets ( Figure S1). An analysis of the composite difference between the high and low solar activity years yields very similar results ( Figure S2).
Because the strongest precipitation response is located in the MYR region, regionally averaged late winter precipitation in the MYR (110-118°E, 26-31°N) is then chosen for further investigation. As shown in Figure 2a, the correlation coefficient between the SSN and late winter MYR precipitation for 1901-2010 is 0.36 (p < 0.01). We also calculated this correlation using in situ precipitation data for 1951-2010 and found that the correlation reached 0.41 (p < 0.01). To remove the noise from a higher-frequency climate variability, the signals with periods in the 9-to 13-year band in rainfall and SSN were filtered out (Figure 2b). The decadal component of precipitation is generally coherent with the SSN. The highest correlation coefficient between these two series, r = 0.75, is obtained at zero lag when the significance is at the 99% level using the Monte Carlo test (Figure 1b). In addition, the wavelet spectrum of the MYR precipitation shows a remarkable quasi-11-year oscillation, which is close to the SSN period, although this spectrum only passes the 90% confidence level (Figure 2c). In general, the results in Figure 2 provide evidence for a significant inphase relationship between the SC and late winter MYR precipitation.
Interestingly, this in-phase relationship is not only statistically significant but also persists through 10 consecutive SCs (Figure 2b), implying a robust SC impact on MYR precipitation. To further examine this, Figure 3a shows the 33-year sliding late winter MYR precipitation response estimated from the MLR analysis over the 1901-2010 period. Notably, the solar index in Figure 3a (i.e., SSN) had been employed at different lead/lag times between 0 and 5 years. The year in the x axis is labeled according to the central year of the 33-year window, while the year of the y axis denotes the lead/lag time employed. As shown in Figure 3a, the precipitation response exhibits a stable synchronous nature, with a positive SC signal generally maximizing at zero lag for most of the time intervals during 1901-2010, while a negative SC signal appears to be the strongest at an~5-year lead/lag. Significant (p < 0.05) signals were mainly detected during the 1970s to 1990s. The lead/lag of 0 (5) years in Figure 3a indicates the maximum (minimum) year of the 11-year SC; hence, the results are consistent with those shown in Figure 2. The same analysis is also performed using GPCC data for 1891-2016 and UDEL data for 1901-2016 ( Figure S3), both of whichshow similar results to those in Figure 3a. To further assess the robustness of the SC-precipitation relationship, the SC signals in the late winter precipitation were calculated using two subseries of the precipitation record spanning the first and second halves of the study period (1901-1955 and 1956-2010, respectively). The precipitation response maintains its pattern during the two subperiods, although it is less statistically significant during the first period.
For comparison, we also performed the same analysis as that shown in Figures 3a-3c for OND, which is shown in Figures 3d-3f. Figure 3d suggests that the SC-precipitation relationship over the MYR region in OND exhibits a changing nature. Before the 1940s, the strongest positive (negative) response appeared at 2-3 years leading (lagging) the SC. However, after the 1940s, the responses are nearly the opposite. At zero lag, negative responses persist in time but are not statistically significant. Precipitation deficits exist in both subperiods over southern China; however, the area with statistical significance is relatively small. In general, the zero lag solar-precipitation relationship over south China is nearly the opposite in early winter from that in late winter, but the early winter relationship seems less robust and significant.

Discussion
As revealed by the above analysis, the in-phase correlation between the SC and MYR precipitation is not only statistically significant but also consistent over time. To further explore the physical mechanisms behind this correlation, we show the SC-associated circulation changes in Figure 4. Figure 4a displays the SC signal at the 500-hPa geopotential height (GPH). Over Asia, positive GPH anomalies occupy Japan, implying a weakening of the East Asian trough. Negative GPH anomalies dominate a vast area from Siberia to the Tibetan Plateau, indicating a deepening of the trough embedded in the westerlies upstream of the CEC region. Correspondingly, anomalous southerly winds, low-level convergence, and midlevel ascending motion appear ahead of the anomalous trough over the CEC region (see Figure 4b). Moreover, these wind anomalies also substantially enhance the moisture transport from the Indian Ocean and the South China Sea to southern China, leading to moisture convergence and thus favoring an increase in the southern China precipitation (see Figure 4c).
Notably, the circulation anomaly over Asia is not a local phenomenon but part of a wave train originating in the high latitudes of the North Atlantic, which is composed of negative GPH centers over Iceland, Middle East and Tibetan Plateau, as well as positive centers over the Mediterranean and Japan. The upper-level wave activity fluxes propagate southeastward from the northern North Atlantic, pass through Europe and northeast Africa, then turn eastward and finally propagate into East Asia, generally consistent with the wave-like GPH anomalies. By performing an MLR analysis similar to that performed in our study, Brugnara et al. (2013;B2013 hereafter) also identified a robust pattern of SC signal in a reconstructed Figure 3. (a) The 33-year sliding SC signals in JFM precipitation over the MYR region at various lags (from −5 to +5 years). The x axis, year, is labeled according to the central year of the 33-year window, which means that at a given year N, the multiple linear regression analysis is performed using data sets from year N − 16 to year N + 16. The y axis shows the number of years that the precipitation data lagging the SSN index in multiple linear regression equation (1). Solid black (white) dots denote that the SC signals are statistically significant at the 5% (1%) level (i.e., p < 0.05 (p < 0.01)) after prewhitening and a 1,000-trial bootstrap resampling test. (b and c) The same as Figure 1c but for SC signals in precipitation during 1901-1955 and 1956-2010, respectively. (d-f) The same as (a)-(c) but for SC signals in OND precipitation. The units of SC signals in Figure 3 are millimeters per month. The definition of the SC signal here is the same as upper air GPH fields during late the winters (JFM) spanning 1927-2002. Since the GPH reconstructions used by these researchers are based on observed upper air data, the B2013 analysis is mainly restricted in the Euro-Atlantic region due to the limited spatial coverage of observations in the early twentieth century. Surprisingly, their derived GPH response pattern is very consistent with the one obtained from the ERA20C reanalysis in this study, as both exhibit a southeast oriented wave-like pattern that include cyclonic anomalies over Iceland. Such consensus increases our confidence in the reliability of the obtained circulation responses.
Based on an empirical orthogonal function analysis of monthly meridional winds at 250 hPa, Hu et al. (2018) identified a wave-like atmospheric teleconnection pattern along the wintertime Asian jet. The wave energy originates from the North Atlantic and propagates to East Asia along the subtropical jet stream waveguide. The researchers found that this pattern can exert significant influences on the rainfall anomalies in South and East China. Notably, the middle-to-upper troposphere GPH and wave activity anomalies associated with this teleconnection pattern (see their Figure 3) are very similar to those associated with the solar pattern derived in this study, further supporting our argument that the SC may exert impacts on MYR precipitation through teleconnections from the Euro-Atlantic region.
As noted in Figure 1, the significant in-phase correlation between the SC and southern China precipitation only exists in late winter, while in early winter, these correlations are nonsignificantly anticorrelated. This change is consistent with the evident subseasonal variations in SC signals in NAE circulation anomalies identified by recent studies (Gray et al., 2016;Ma et al., 2018). In late winter, the circulation response features a significant negative pressure anomaly over the Iceland region that is approximately synchronous ). Note that, in (b) and (c), solid white dots denote 90% confidence level for vertical velocity and moisture divergence, only the wind vector and moisture fluxes significant at the 90 % confidence level are shown. The definition of the SC signal here is the same as Figure 1. to the SC, which reflects a direct top-down influence from the stratosphere via heating effects of the solar ultraviolet irradiance (Chen et al., 2015;Ineson et al., 2011). Then, the Icelandic pressure anomaly can affect southern China precipitation through Rossby wave trains. However, in early winter (see Figure S4), such a west-east oriented wave train response does not exist. Instead, there is a negative GPH anomaly over Scandinavia and a positive GPH anomaly over central Asia (see Figure S4a), implying a strengthening of the ridge upstream of southern China. As a result, anomalous midlevel descending motion appears ahead of the anomalous ridge (see Figure S4b). These circulation features favor a decrease in precipitation over southern China. However, the strength and significance of the early winter circulation and moisture transport (see Figure S4c) response are relatively weak; thus, the precipitation response is less significant.

Concluding Remarks
In southern China, severe wintertime rainfall/snowstorm events occasionally occur, causing catastrophic consequences to society and the economy. Thus, increasing attention has recently been devoted to the causes of winter precipitation anomalies over southern China. As an important source of regional decadal climate variability during the Northern Hemisphere winter, the SC role in winter precipitation variability over southern China remains unclear. In this study, CRU land surface precipitation and SSN data were analyzed with the MLR approach to provide evidence of a connection between them.
The results suggest an evident subseasonal evolution of the SC-precipitation relationship; in the early winter months, only weak and nonsignificant negative precipitation responses can be observed in southern China. In late winter, the signal exhibits a region of enhanced precipitation centered over the MYR basin that extends across most of southern China, and the response over the MYR basin is statistically significant at the 99% confidence level. Further analysis suggests that the in-phase correlation between SC and late winter MYR precipitation is not only statistically significant but also consistent over time. The precipitation responses result from the solar-associated Rossby wave trains propagating from the NAE region, and the responses are composed of negative GPH centers over Iceland, the Middle East, and the Tibetan Plateau, as well as positive centers over the Mediterranean and Japan. Such a circulation configuration suggests weakening of the East Asian trough and a deepening of the trough upstream of southern China, leading to moisture convergence and thus favoring an increase in precipitation over southern China.
Finally, it is highlighted that considering the quasiperiodicity of the 11-year sunspot cycle, our results may potentially help to improve decadal predictions of late winter precipitation over southern China, particularly over the MYR region.