Impact of late spring Siberian snow on summer rainfall in South-Central China

Located in the Yangtze River Valley and surrounded by mountains, South-Central China (SCC) frequently suffered from natural disasters such as torrential precipitation, landslide and debris flow. Here we provide corroborative evidence for a link between the late spring (May) snow water equivalent (SWE) over Siberia and the summer (July–August, abbr. JA) rainfall in SCC. We show that, in May, anomalously low SWE over Siberia is robustly related to a large warming from the surface to the mid-troposphere, and to a stationary Rossby wave train from Siberia eastward toward the North Atlantic. On the one hand, over the North Atlantic there exhibits a tripole pattern response of sea surface temperature anomalies in May. It persists to some extent in JA and in turn triggers a wave train propagating downstream across Eurasia and along the Asian jet, as the so-called Silk Road pattern (SRP). On the other hand, over northern Siberia the drier soil occurs in JA, accompanied by an overlying anomalous anticyclone through the positive feedback. This anomalous anticyclone favors the tropospheric cooling over southern Siberia, and the meridional (northward) displacement of the Asian jet (JMD) due to the change in the meridional temperature gradient. The combination of the SRP and the JMD facilitates less water vapor transport from the tropical oceans and anomalous descending motion over SCC, and thus suppresses the precipitation. These findings indicate that May Siberian SWE can be exploited for seasonal predictability of SCC precipitation.


Introduction
The mountainous areas drained by the Yangtze River and its tributaries (i.e., Yangtze River Valley, abbr. YRV) are regions of rapid economic development and population growth at great risk from natural disasters. Particularly, South-Central China (SCC) is highly susceptible to extreme flooding and drought events. For example, SCC suffered from the extreme drought and heat wave of the summer 2013, which affected about 47.8 million people, 3.63 million livestock and 48 thousand km 2 arable land and caused direct economic losses up to ¥30 billion (Duan et al. 2013).
Previous studies demonstrated that the sea surface temperature anomalies (SSTAs) over the tropical Pacific Ocean, the tropical Indian Ocean as well as over the north Atlantic Ocean, give rise to changes in summer rainfall over SCC. Traditionally, the strongly coupled sea-air interactions in the tropics, known as the El Niño-Southern Oscillation (ENSO), is a notable external forcing of the summer rainfall variability over SCC, in which a long-maintained, Electronic supplementary material The online version of this article (https ://doi.org/10.1007/s0038 2-020-05206 -5) contains supplementary material, which is available to authorized users.
1 3 lower-tropospheric anticyclone over the Philippines causes more moisture transport along its western boundary (Wang et al. 2000;Huang et al. 2004). It is noteworthy that the relationship between ENSO and the summer rainfall over SCC is not stable on the multidecadal timescales (Wang 2002). Over the analysis period of 1964-1995in Wang (2002, significant correlation emerges only during 1964-1974 and 1983-1990, when there is large interannual variability of the low-level temperature and of the subtropical high over the tropical Pacific. Additionally, Shen et al. (2019) found the reverse August precipitation anomaly over eastern China in 1998 and 2016, which are both the super El Niño events in history. Early studies also emphasized that the tropical Indian Ocean (TIO) SST acts as a capacitor, anchoring the suppressed convection and lower-tropospheric anticyclone over the Philippines during the El Niño decay phase (Yang et al. 2007;Xie et al. 2009). Moreover, it has been documented that the tripole pattern of the North Atlantic SSTAs and the phase of the North Atlantic Oscillation (NAO) modulate the summer rainfall variability over SCC, via triggering a stationary Rossby wave train extending from the North Atlantic toward East Asia (Sung et al. 2006;Linderholm et al. 2011;Tian and Fan 2012). Land surface conditions, such as soil moisture, also influence the summer rainfall variability over SCC (Zhang and Zuo 2011;Meng et al. 2014). Drier soil conditions in spring, stretching from the Yangtze River valley to North China, increase the surface air temperature and hence strengthen the East Asian summer monsoon (EASM) and summer rainfall over SCC by enhancing the sea-land thermal difference (Zhang and Zuo 2011). Halder and Dirmeyer (2017) demonstrated that negative soil moisture anomalies over eastern Eurasia in spring induce an anomalous upper-tropospheric ridge around 100° E via anomalous surface and mid-tropospheric heating, which further modulates the Asian jet and summer rainfall over Asia.
Snow is another important land surface factor that exerts a strong control on the overlying atmosphere and even on the hemispheric-scale circulation. Via the radiative snow-albedo feedback and the thermodynamical feedback (the insulating snow layer decoupling the lower atmosphere from the soil), a thicker snowpack (or higher snow water equivalent, SWE) cools the lowermost atmosphere (Walsh et al. 1985;Groisman et al. 1994). In addition, there is the hydrological feedback whereby positive (negative) snow anomalies convert in positive (negative) soil moisture anomalies with a delay, in the melting season. Previous studies demonstrated that an anomalous Siberian snow cover can be accompanied by polar vortex and northern annular mode anomalies during autumn and winter, coupling the troposphere to the stratosphere (Cohen et al. 2007;Fletcher et al. 2009;Henderson et al. 2018). The snowpack can have significant impacts on the atmospheric circulation not only during the contemporaneous season but also in the following seasons. Based upon the observational analyses, many studies found a negative correlation between the spring or summer Siberian snow cover/depth and the strength of the Indian summer monsoon through the combination of radiative, thermodynamical and hydrological feedbacks (e.g. Hahn and Shukla 1976;Dickson 1984;Bamzai and Shukla 1999;Fasullo 2004;Dash et al. 2005). The impact and the mechanism have also been revealed by the numerical models (Yasunari et al. 1991;Bamzai and Marx 2000;Dash et al. 2006).
Some studies specifically explored the impact of the Siberian snow on the East Asian summer rainfall. Based on the singular value decomposition (SVD) analysis, the SWE over Eurasia during spring derived from National Snow and Ice Center (NSIDC) has been linked to the summer rainfall in China during the period of 1979-2004). By using the empirical orthogonal function (EOF) analysis for snow cover data from NOAA satellites, Yim et al. (2010) noted that the east-west dipole mode of snow cover anomalies (with the positive and negative values over western and eastern Eurasia, respectively) is closely related to the EASM during 1972-2004. But the significant summer rainfall anomalies associated with this dipole were only observed over Korea and Japan, not over eastern China. Analyzing the period of 1979-2013, Zhang et al. (2017) found that the east-west dipole mode of the spring SWE decrement (SWE in February minus SWE in May) is associated with the summer rainfall over East Asia through the local Eurasian soil conditions persisting from spring to summer. However, Robock et al. (2003) argued that the soil moisture alone could not explain the impact of the preceding Eurasian snow on the summer precipitation over Asia. While a moderate relationship between SWE over Siberia in May and rainfall over China in summer was found in reanalysis and seasonal reforecasts during 1983-2010 (Zuo et al. 2015), the linking mechanism was not elucidated. Though the Siberian snow is suggested to have a significant impact on the Asian summer climate, these studies have suffered from several limitations.
(1) The analysis period is relatively short. Most previous studies only covered the period ending in the early twentyfirst century. (2) There is uncertainty in the snow observations. For example, snow cover is derived from optical and infrared remote sensing by the NOAA satellites, and there is uncertainty associated with the conversion of binary pixel information about snow cover to large-scale snow cover gridded data. On the other hand, the widely used SWE data provided by the NSIDC is derived from microwave remote sensing, and there is inaccuracy due to the retrieval method using a static algorithm (Xu et al 2018); (3) Although Zhang et al. (2017) pointed out the importance of the snow persistent influence into the summer through the hydrological feedback for maintaining eastward-propagating wave trains across Eurasia, the relative roles of SST and land conditions have not been fully clarified.
The present study investigates the potential linkage between late spring (May) SWE over Siberia and summer (July-August) rainfall in SCC for the period 1979-2018, based on SWE and soil moisture data retrieved from the ECMWF (European Centre for Medium-Range Weather Forecasts) Interim/Land reanalysis (with more information in Sect. 2). The important roles played by the North Atlantic SST and Siberian soil moisture to perdure the influence of the spring Siberian snow into the summer season and to connect the latter with precipitation over SCC are explored quantitatively.

Data, climatic indices and methods
This study utilizes five datasets. The monthly (1) SWE and (2) soil moisture in three layers (7 cm, 21 cm, 72 cm) are obtained from the ERA-Interim/Land with a resolution of 1° × 1° (Balsamo et al. 2015). (3) The monthly and daily atmospheric fields are collected from the ERA-Interim reanalysis, with a horizontal resolution of 1° × 1° (Dee et al. 2011). (4) The monthly precipitation data are retrieved out of the monthly mean CPC Merged Analysis of Precipitation (CMAP), which are available in a 2.5° × 2.5° grid (Xie and Arkin 1997). (5) The monthly SST data are provided by the Met Office Hadley Centre (Rayner et al. 2003), with a resolution of 1° × 1°. The analyzed period in this study covers from 1979 to 2018.
The ERA-interim/land snow data is a high spatial resolution reanalysis driven by realistic meteorological forcing. Wegmann et al. (2017) has validated the ERAinterim/land reanalysis against the in-situ station data over northern Russia. The Taylor diagram (their Fig. 5) displays the daily variability of snow depth in ERA-interim/land against the in-situ observation over 13 Russian stations over the period 1981-2010, in which their correlation is 0.8 in April and their standard deviations are comparable. Moreover, we validate the SWE data in the ERA-interim/ land against the relatively long-period, satellite-based SWE dataset from the Finnish Meteorological Institute (FMI), with a spatial resolution of 25 km from 1979 to 2014 (Takala et al. 2011; see detailed information at http:// www.globs now.info/). The SWE product from FMI combines satellite-based passive microwave measurements with ground-based weather station data in a data assimilation scheme. For the interannual variability, the ERAinterim/land data is highly consistent with the FMI data over Siberia (Fig. S1a). The spatial distributions of the SWE climatology from the FMI data and the ERA-interim/ land data are quite similar, though the ERA-interim/land data overestimate the magnitude of the SWE over the Central and East Siberian Plateaus (Fig. S1b). Taken together, it confirms that the ERA-interim/land reanalysis is an appropriate dataset that can be used in this study.
The definitions of the climatic indices are given in Table 1. All indices are standardized. To isolate the influence of Siberian snow on the atmospheric circulation and precipitation at the interannual timescale, any linear trend has been removed prior to analysis from all the indices and fields. The statistical methods used in the current study include correlation analysis, linear regression and SVD analysis. The statistical significance of correlation and regression is assessed using the two-tailed Student's t test.
To illustrate the wave-like activity, the wave activity flux (WAF) is applied in the study (Takaya and Nakamura 2001). In order to diagnose the excitation of Rossby waves, the wave source term defined as −∇ ⋅ ⃗ V x (f + ) (Sardeshmukh and Hoskins 1988) is calculated, where ⃗ V x is the divergent wind velocity, f is the Coriolis parameter, and is the relative vorticity. The Siberian SWE (50°-75° N, 60°-140° E) in May is emphasized in this study with the largest interannual variability (Fig. S2). Additionally, the Siberian snow melts a lot in May, except for a few regions at very high altitudes or along the Arctic coast (Xu and Dirmeyer 2013). The spatial distribution of precipitation variations over China in June and July-August are distinct: the largest variability is located 1 3 over South China Sea and over Yangtze River Valley, respectively (Wang et al. 2009, their Fig. 4). Hence, in the current study, we focus on the precipitation during the late summer (July-August) rather than the 3-month (June-August) mean. Figure 1 illustrates the leading SVD mode for the May SWE over Siberia and the July-August (JA) precipitation over eastern China. The leading mode accounts for 18.8% of the total interannual variance of the Siberian SWE anomalies in May. Notable are negative SWE anomalies over Siberia in May, especially over central and eastern regions (Fig. 1a). Meanwhile, there are below-normal precipitation in JA over parts of the Inner Mongolia and YRV, particularly over SCC (Fig. 1b). The corresponding time series (Fig. 1c) indicates a statistically significant linkage between Siberian SWE in May and summer precipitation over SCC, with a coefficient of 0.81 (above the 99% confidence level). Here we define the SWE index (SWEI) using the normalized time series of the SWE variations in the leading SVD mode (positive SWEI corresponds to reduced SWE over Siberia). The area-averaged precipitation anomalies over SCC (the frame marked in Fig. 1b), multiplied by − 1, is taken as the precipitation index (PI), implying that a positive value indicates below-normal precipitation. As expected, the correlation coefficient between SWEI and PI is 0.48 ( Fig. 2a; above 99% confidence level). It is noteworthy that these results can be reproduced by using the SWE data from the FMI (Fig. S3). Figure 2b illustrates the water vapor flux anomalies integrated vertically from 1000 to 300 hPa in JA regressed onto the SWEI. The water fluxes are indicative of an anomalous anticyclone over the western North Pacific (35° N), implying a westward-extended western Pacific subtropical high ( Fig. 2b: vectors). At lower latitudes (20° N), around 120° E, they are also indicative of an anomalous cyclonic circulation over southeastern China, with the northerly flow decreasing the water vapor flux from the tropical oceans to SCC (Fig. 2b: vectors), resulting in significantly positive divergence anomalies in SCC ( Fig. 2b: shading and frame). Besides, the meridional-vertical cross section of vertical velocity anomalies regressed upon the SWEI, averaged between 105°-120° E (Fig. 2c), shows anomalous descending motion around 25°-32° N throughout the with a positive value indicating the snow/precipitation decrease. In b, the purple curves denote the Yangtze River and Yellow River, respectively, and the region marked by dashed lines denotes the South-Central China here and hereafter entire troposphere. Taken together, less SWE over Siberia in May is robustly linked to positive water vapor flux divergence anomalies and anomalous descending motion, which suppresses the summer precipitation over SCC.

The influences of the preceding Siberian SWE in May
Previous studies have revealed that variation of snow conditions has an impact at the surface and in the troposphere via radiative, hydrological and thermodynamical effects (e.g. Barnett et al. 1989;Cohen and Rind 1991;Dash et al. 2005;Sun 2017). Figure 3a shows the SWE, tropospheric air temperature and zonal wind anomalies in May along the 120° E meridian regressed onto the SWEI.
Corresponding to the significantly negative SWE anomalies over Siberia (between 60° N and 70° N), there is a significant warm-core in the lower-troposphere ( Fig. 3a: shading), which may attribute to a positive surface sensible heat flux anomaly over Siberia (Fig. S4a). Besides, an anomalous anticyclone is apparent over the northern Siberia-North Pacific Sector through the snow-atmospheric coupling (Figs. S4b and 4c). Negative and positive zonal wind anomalies emerge in the southern and northern flanks of the anomalous warm-core anticyclone ( Fig. 3a; contours). Xu and Dirmeyer (2011) has revealed the strong snow-lower atmosphere coupling over Siberia in May, and Xu and Dirmeyer (2013) further demonstrated that the vertical extent of this coupling is up to the mid-troposphere (500 hPa). Figure 3b illustrates the geopotential height and horizontal WAF anomalies at 300 hPa in May regressed onto the SWEI. A largely positive geopotential height anomaly at 300 hPa is found over the Siberia-North Pacific sector in May, related to the reduced SWE, together with alternating negative and positive height anomalies downstream ( Fig. 3b; contours). This signature is consistent with the formation of an apparent Rossby wave train stretching from the eastern North Pacific to western North America. It then ramifies into two branches: one propagates southward toward the lower latitudes, and the other extends eastward into the mid-latitude North Atlantic (120°-90° W). Notable is that the latter branch is observed stretching northeastward to Europe, resulting in a negative height anomaly center over western Europe (Fig. 3b: vectors). The aforementioned Rossby wave source (RWS) displays strong positive anomalies over Siberia (Fig. 4a). Previous studies suggested that the advection of vorticity by the divergent and convergent component of the upper tropospheric flow acts as an effective RWS (Sardeshmukh and Hoskins 1988; Chen and Huang 2012). Due to the weakened westerly wind induced by the thermal anomaly (Fig. 3a), the horizontal wind at 300 hPa converges over Siberia, generating a positive RWS anomaly through the  ). b-c Regressions of b vertically-integrated water vapor flux (from 1000 to 300 hPa; vectors; kg m −1 s −1 ) and its divergence (shading; 10 −5 kg m −2 s −1 ) anomalies in JA, and c meridional-vertical cross section averaged along 105°-120° E for the vertical wind (vectors; m s −1 ) and omega (shading; 10 −2 Pa s −1 ) anomalies in JA, upon the SWEI in May during 1979-2018. Data over the Tibetan Plateau in b is masked out and the shape of the Tibetan Plateau is derived from Zhang et al. (2002). Values stippled in b and enclosed by the black contours in c are significant at the 90% confidence level, based on the Student's t test positive vorticity advection by the convergent flow (Fig. 4a,  b). These results indicate that the SWE anomalies over Siberia are associated with eastward-propagating Rossby wave trains to the North Atlantic via the anomalous upper-level divergent flow.
Focusing on the North Atlantic in May, we note the meridionally banded structures of the zonal wind anomalies (Fig. 5a), with two bands negative anomalies around 45° N and 25° N and one band of positive anomaly between them. The results suggest the deceleration of both the eddydriven (45° N) and subtropical (25° N) jets over the North Atlantic. The decelerated jets induce the easterly and southerly wind anomalies near the surface, and further lead to the northern (50°-60° N, 30°-60° W) positive centers of the SSTAs (Fig. 5b: shading). In addition, the decelerated westerly jet may also reinforce the meridional anomalies of the atmosphere. It appears the northerly wind anomaly to the south (30°-40° N, 40°-70° W), weakening the warming current from the Gulf Stream (Rossby 1996), which cools the underlying SST. The tripole pattern of SSTAs is thus apparent. Moreover, there is an in-phase relationship between the turbulent heat flux anomalies and the SSTAs, especially over the two SST anomaly centers: the negative SSTAs around 30°-40° N concur with the negative turbulent heat flux anomalies, and the positive SSTAs around 50°-60° N concur with the positive turbulent heat flux anomalies (Fig. 5b: contours). Positive turbulent heat flux anomalies represent downward flux, and this relation suggests that May SSTAs over the North Atlantic are mainly instigated by the atmosphere (See vectors in Fig. 5b). In conclusion, less Siberian contours; m s −1 ) and the underlying SWE (lower panel; mm) anomalies in May, and b geopotential height (contours; gpm) and associated horizontal wave activity fluxes (WAFs; vectors; m 2 s −2 ) anomalies at 300 hPa in May, upon the SWEI in May during 1979-2018. Con-tours in a are at ± 0.4, ± 0.8, ± 1.2 m s −1 and in b are at ± 4, ± 8, …, ± 28 gpm. The thick blue arrows in b denote the Rossby wave trains. Stippled regions and green markers in a, and shading in b denote the values are significant at the 90% confidence level, based on the Student's t test SWE in May is associated with the local tropospheric warming up to 400 hPa, which induces a Rossby wave train propagating eastward toward the North Atlantic. The associated westerly jets over the North Atlantic decelerate, which further contributes to the tripole pattern of Atlantic SSTAs.

The connecting roles of the North Atlantic SST
and Siberian soil moisture in JA Figure 6a illustrates the SST and turbulent heat flux anomalies in JA regressed onto the SWEI. In comparison with Fig. 5b, the two anomaly centers of the SST over the mid-to high-latitude North Atlantic persist. However, the significant anomaly center over the low-latitude North Atlantic northwestward shifts around the Gulf Stream. More interestingly, the relationship between the anomalous turbulent heat flux and the SSTAs becomes out-of-phase, especially over the two anomaly centers around 35° N and 45° N (the black frames in Fig. 6a). This out-of-phase relationship indicates that the SSTAs exert an influence onto the overlying atmosphere. Here the area-averaged SST in these two anomaly centers is used to define the North Atlantic SST index (SSTI), and its correlation coefficient with the SWEI is 0.34 (above 95% confidence level; Fig. 6b). Figure 6c shows the 200 hPa horizontal wind anomalies in JA regressed onto the SSTI. There is a largely anomalous cyclone over the North Atlantic, accompanied with cyclonic and anticyclonic circulation anomalies downstream across Eurasia (Fig. 6c: vectors). Besides, the apparent positive and negative anomalies of the 200 hPa meridional wind indicate a wave train from the North Atlantic eastward to Eurasia (Fig. 6c: contours). From the North Atlantic, there are two branches of the Rossby wave train over Eurasia: the Ural-Siberia route (northern branch) and the Mediterranean-East Asia route (southern branch; e.g. Orsolini et al., 2015). The latter one along the southern slope of the Tibetan Plateau and the climatological jet axis resembles the so-called Silk Road pattern (SRP; Lu et al. 2002;Hong and Lu 2016). The SRP index (see the definition in Table 1, referring to Chen and Huang 2012) is highly correlated with the SSTI, with a coefficient of 0.48 (above 99% confidence level, Table 2). Consistently, previous studies have pointed out that the SSTAs over the North Atlantic can excite the stationary wave train to East Asia (Gu et al. 2009;Tian and Fan 2012). Further, we focus on the connecting role of the Siberian soil moisture. Figure 7a shows the top meter soil moisture anomalies in JA regressed onto the SWEI. A largely negative soil moisture anomaly occurs in most parts of Siberia, primarily because of less May Siberian SWE into June, but mainly confined to north of 65° N (Fig. S5). Hence, less SWE in May-June leads to less local melting in JA. We further define a soil moisture index based upon this negative center (SMI; multiplied by − 1, hence a positive value implying the drier soil moisture; Fig. 7b). The correlation of SMI and SWEI is 0.36 (Fig. 7b). There are also other localized positive (90° E and 130° E) and negative (65° E and 105° E) soil moisture anomalies alternatively over southern Siberia, which may be induced by the alternating anticyclonic and cyclonic anomalies along the Ural-Siberia wave route (Figs. 6c, 8b). Figure 7c illustrates the temperature advection at 850 hPa, the vertical integral of temperature from 1000 to 200 hPa and its meridional gradient in JA regressed Contours in a are at ± 2, ± 4, ± 6 × 10 5 J m −2 and in c are at ± 0.5, ± 1.0, ± 1.5 gpm onto the SMI. When conditions of local drier soil prevail, the tropospheric temperature increases over the northern Siberia and East Asia-North Pacific sector and reduces over southern Siberia (Fig. 7c: contours). The anomalous temperature anomalies are associated with the cold and warm advections ( Fig. 7c: vectors), which concur with the anticyclonic anomalies over northern Siberia through the positive feedback with the underlying drier soil ( Fig. S6; Fischer et al. 2007). Consequently, negative temperature gradient anomalies are observed over 60° N and 30° N, in conjunction with positive anomalies over 45° N (Fig. 7c: shading). The zonal wind anomalies at 300 hPa regressed onto the SMI in Fig. 7d exhibits weakened westerly winds over 60° N and 30° N, and strengthened westerly winds over 45° N, which are consistent with the meridional temperature gradient anomalies. These upper-level zonal wind anomalies around the climatological Asian jet axis indicate a meridional (northward) displacement of the Asian jet (JMD; Lin and Lu 2005;Hong and Lu 2016). The JMD index (JMDI) is closely related to the SMI, and their correlation coefficient is 0.41 (above 99% confidence level, Fig. 7b). Both the North Atlantic SSTAs and the Siberian soil moisture anomalies in JA are closely related to the SSC precipitation in JA. Based on the partial correlation coefficients, the relative contributions of the SST and the soil moisture to the SSC precipitation in JA are 5.8% and 16.8%, respectively. We thus define a Sea surface Temperature-soil Moisture index (STMI) in JA (Fig. 8a) [SSTI, SWEI]) means the correlation coefficient between SMI (SSTI) and SWEI. Figure 8b depicts the 200 hPa zonal and meridional wind anomalies regressed onto the STMI. In terms of the meridional wind, the wave train closely resembles the regression onto the SSTI (Fig. 6c), except over Eurasia where the SRP along the jet becomes much stronger and more significant. For the zonal wind (Fig. 8b: shading), it reproduces the northward JMD, consistent with the regression onto the SMI (Fig. 7d). Therefore, the effects of the preceding SWE in May can be well represented by the STMI, which involves both the SRP and JMD, referred to as the upstream and downstream effects on SCC precipitation, respectively. Figure 9 depicts the precipitation anomalies in JA regressed onto the PI, JMDI, SRPI and STMI. As expected, the JMDI-, SRPI-and STMI-regression patterns closely resemble that regressed onto the PI. All the three patterns display an apparently negative center of JA precipitation over SSC, although relatively weaker compared to the PIrelated pattern. Furthermore, as shown in Fig. S7, the SRP is associated with the decreased water vapor primarily due to meridional wind anomalies. Meanwhile, the northward JMD regulates anomalous descending motion over SSC (also see Fig. 2). In addition, the regressions onto the SSTI and SMI similarly show a negative anomaly center over SCC, despite their weaker intensity (Fig. S8). It's notable that the precipitation anomalies over Inner Mongolia are also correlated with the Siberian SWE (Figs. 1b, 10e), which hasn't been discussed in this paper. Interestingly, the precipitation anomalies associated with the SSTI also show similar but weaker anomalies over Inner Mongolia (Figs. S8a and S8b). It suggests that the North Atlantic SSTAs may influence the precipitation anomalies over Inner Mongolia.

Conclusions and discussion
The previous studies have explored the relationship between Siberian SWE in Spring and the East Asian precipitation in summer Zhang et al. 2017). However, in this study, we emphasize that the SST over the North Atlantic and the Siberian soil moisture have play important linking roles in the Siberian SWE-SCC precipitation connection. These physical processes can be described schematically as follows (also see Fig. S9): • Corresponding to the below-normal Siberian SWE anomalies in May (Fig. 10a), over Siberia there is significant tropospheric warming from the surface into 400 hPa and largely positive 300 hPa geopotential height anomaly. It instigates the Rossby wave train originated over Siberia and propagating eastward across Pacific and toward the North Atlantic (Figs. 3, 4). The associated Atlantic jet weakens, following by a tripole pattern response of SSTAs over the North Atlantic (Fig. 5). It is noteworthy that, on the one hand, the May tripole pattern of SSTAs over the North Atlantic persists into JA to some extent (Fig. 10b), and on the other hand, the Siberian soil moisture in JA is drier-than-normal owing to less Siberian SWE in May-June (Fig. 10c). The time series of the SWEI in May (red solid line), the soil moisture index (SMI) in JA (coral dash line) and the JMDI in JA (black dash line). c-d Regressions of c temperature advection at 850 hPa (vectors; K m s −1 ), vertically-integrated temperature (from 1000 to 200 hPa; contours; K) and its meridional gradient (shading; 10 −2 K m −1 ) anomalies in JA, and d zonal wind anomalies at 300 hPa (shading; m s −1 ) in JA, upon the SMI in JA during 1979-2018. Contours in c are at ± 1, ± 2, ± 3, ± 4 × 10 4 K. The purple (green) vectors in c delineate warm (cold) air advections, with only magnitude larger than 100 shown. Stippled values in a, c, d are significant at the 90% confidence level, based on the Student's t test ◂ • In JA, the SSTAs over the North Atlantic in turn may excite a Rossby wave train, referred as the SRP. It is characterized by the alternating northerly and southerly wind anomalies from the North Atlantic, along the southern slope of the Tibetan plateau and toward East Asia (Fig. 6). In addition, the drier Siberian soil concurs with an anomalous overlying anticyclone through the positive feedback (Fischer et al. 2007; Fig. S7). The associated cold air advection along the eastern and southern flanks of the anomalous anticyclone cools the tropospheric temperature over southern Siberia, which further favors the JMD by changing in the meridional temperature gradient (Fig. 7). • The combination of the SRP and the JMD (i.e., the upstream and downstream effects, respectively; Figs. 8, 10d) contributes to less water vapor transport from the tropical ocean and anomalous descending motions around 25°-32° N, and hence the lack of precipitation in SCC (Figs. 2, 10e). Fan et al. (2008) proposed a statistical model that can explain 64% of the interannual variability of the YRV summer precipitation, based on six predictors (Antarctic Oscillation, Ural circulation, East Asia circulation, meridional wind shear, South Pacific circulation, and low-level vorticity). The hybrid downscaling models, based on the simultaneous predictors from general circulation models (500 hPa geopotential height and 850 hPa specific humidity) and the preceding predictors from the reanalysis data (700 hPa geopotential height and sea level pressure), are also applied for the summer precipitation prediction over China (Liu and Fan  Fig. 10 Schematic diagram summarizing the dynamical linkage between the snow depth decrease over Siberia in May with the precipitation anomalies over SCC in JA. a Spatial distributions of SWE in May over Siberia of the leading SVD mode (as Fig. 1a); Regressions of b the SST anomalies in May (shading; K) and JA (contours; K); c the snow depth anomalies in June (contours; mm) and the top meter soil moisture anomalies in JA (shading; 10 −2 m 3 m −3 ); d the zonal wind anomalies (shading; m s −1 ) and meridional wind anoma-lies (contours; m s −1 ) at 200 hPa in JA; and e JA precipitation anomalies (shading; mm day −1 ) in China east of 100° E, upon the SWEI in May during 1979-2018. Contours in b are at ± 0.05, ± 0.1, ± 0.15, ± 0.2 °C, in (c) are at − 32, − 22, − 12, − 2 mm and in d are at ± 0.4, ± 0.8, ± 1.2 m s −1 . The thick red line in c denotes the 90% confidence level of the SWE anomalies in June. Stippled values are significant at the 90% confidence level, based on the Student's t test 2014). However, these prediction models did not consider the effects of the preceding Siberian snowpack. Our study indicates that the May Siberian snow is closely related to the summer precipitation over SCC, and explains 23% of the total variance. Considering May SWE as a potential predictor can help improving the summer SCC precipitation predictability.